Search icon CANCEL
Subscription
0
Cart icon
Your Cart (0 item)
Close icon
You have no products in your basket yet
Save more on your purchases! discount-offer-chevron-icon
Savings automatically calculated. No voucher code required.
Arrow left icon
Explore Products
Best Sellers
New Releases
Books
Events
Videos
Audiobooks
Packt Hub
Free Learning
Arrow right icon
timer SALE ENDS IN
0 Days
:
00 Hours
:
00 Minutes
:
00 Seconds

How-To Tutorials

7018 Articles
article-image-building-recommendation-system-azure
Packt
19 Feb 2016
7 min read
Save for later

Building A Recommendation System with Azure

Packt
19 Feb 2016
7 min read
Recommender systems are common these days. You may not have noticed, but you might already be a user or receiver of such a system somewhere. Most of the well-performing e-commerce platforms use recommendation systems to recommend items to their users. When you see on the Amazon website that a book is recommended to you based on your earlier preferences, purchases, and browse history, Amazon is actually using such a recommendation system. Similarly, Netflix uses its recommendation system to suggest movies for you. (For more resources related to this topic, see here.) A recommender or recommendation system is used to recommend a product or information often based on user characteristics, preferences, history, and so on. So, a recommendation is always personalized. Until recently, it was not so easy or straightforward to build a recommender, but Azure ML makes it really easy to build one as long as you have your data ready. This article introduces you to the concept of recommendation systems and also the model available in ML Studio for you to build your own recommender system. It then walks you through the process of building a recommendation system with a simple example. The Matchbox recommender Microsoft has developed a large-scale recommender system based on a probabilistic model (Bayesian) called Matchbox. This model can learn about a user's preferences through observations made on how they rate items, such as movies, content, or other products. Based on those observations, it recommends new items to the users when requested. Matchbox uses the available data for each user in the most efficient way possible. The learning algorithm it uses is designed specifically for big data. However, its main feature is that Matchbox takes advantage of metadata available for both users and items. This means that the things it learns about one user or item can be transferred across to other users or items. You can find more information about the Matchbox model at the Microsoft Research project link. Kinds of recommendations The Matchbox recommender supports the building of four kinds of recommenders, which will include most of the scenarios. Let's take a look at the following list: Rating Prediction: This predicts ratings for a given user and item, for example, if a new movie is released, the system will predict what will be your rating for that movie out of 1-5. Item Recommendation: This recommends items to a given user, for example, Amazon suggests you books or YouTube suggests you videos to watch on its home page (especially when you are logged in). Related Users: This finds users that are related to a given user, for example, LinkedIn suggests people that you can get connected to or Facebook suggests friends to you. Related Items: This finds the items related to a given item, for example, a blog site suggests you related posts when you are reading a blog post. Understanding the recommender modules The Matchbox recommender comes with three components; as you might have guessed, a module each to train, score, and evaluate the data. The modules are described as follows. The train Matchbox recommender This module contains the algorithm and generates the trained algorithm, as shown in the following screenshot: This module takes the values for the following two parameters. The number of traits This value decides how many implicit features (traits) the algorithm will learn about that are related to every user and item. The higher this value, the precise it would be as it would lead to better prediction. Typically, it takes a value in the range of 2 to 20. The number of recommendation algorithm iterations It is the number of times the algorithm iterates over the data. The higher this value, the better would the predictions be. Typically, it takes a value in the range of 1 to 10. The score matchbox recommender This module lets you specify the kind of recommendation and corresponding parameters you want: Rating Prediction Item Prediction Related Users Related Items Let's take a look at the following screenshot: The ML Studio help page for the module provides details of all the corresponding parameters. The evaluate recommender This module takes a test and a scored dataset and generates evaluation metrics, as shown in the following screenshot: It also lets you specify the kind of recommendation, such as the score module and corresponding parameters. Building a recommendation system Now, it would be worthwhile that you learn to build one by yourself. We will build a simple recommender system to recommend restaurants to a given user. ML Studio includes three sample datasets, described as follows: Restaurant customer data: This is a set of metadata about customers, including demographics and preferences, for example, latitude, longitude, interest, and personality. Restaurant feature data: This is a set of metadata about restaurants and their features, such as food type, dining style, and location, for example, placeID, latitude, longitude, price. Restaurant ratings: This contains the ratings given by users to restaurants on a scale of 0 to 2. It contains the columns: userID, placeID, and rating. Now, we will build a recommender that will recommend a given number of restaurants to a user (userID). To build a recommender perform the following steps: Create a new experiment. In the Search box in the modules palette, type Restaurant. The preceding three datasets get listed. Drag them all to the canvas one after another. Drag a Split module and connect it to the output port of the Restaurant ratings module. On the properties section to the right, choose Splitting mode as Recommender Split. Leave the other parameters at their default values. Drag a Project Columns module to the canvas and select the columns: userID, latitude, longitude, interest, and personality. Similarly, drag another Project Columns module and connect it to the Restaurant feature data module and select the columns: placeID, latitude, longitude, price, the_geom_meter, and address, zip. Drag a Train Matchbox Recommender module to the canvas and make connections to the three input ports, as shown in the following screenshot: Drag a Score Matchbox Recommender module to the canvas and make connections to the three input ports and set the property's values, as shown in the following screenshot: Run the experiment and when it gets completed, right-click on the output of the Score Matchbox Recommender module and click on Visualize to explore the scored data. You can note the different restaurants (IDs) recommended as items for a user from the test dataset. The next step is to evaluate the scored prediction. Drag the Evaluate Recommender module to the canvas and connect the second output of the Split module to its first input port and connect the output of the Score Matchbox Recommender module to its second input. Leave the module at its default properties. Run the experiment again and when finished, right-click on the output port of the Evaluate Recommender module and click on Visualize to find the evaluation metric. The evaluation metric Normalized Discounted Cumulative Gain (NDCG) is estimated from the ground truth ratings given in the test set. Its value ranges from 0.0 to 1.0, where 1.0 represents the most ideal ranking of the entities. Summary You started with gaining the basic knowledge about a recommender system. You then understood the Matchbox recommender that comes with ML Studio along with its components. You also explored different kinds of recommendations that you can make with it. Finally, you ended up building a simple recommendation system to recommend restaurants to a given user. For more information on Azure, take a look at the following books also by Packt Publishing: Learning Microsoft Azure (https://www.packtpub.com/networking-and-servers/learning-microsoft-azure) Microsoft Windows Azure Development Cookbook (https://www.packtpub.com/application-development/microsoft-windows-azure-development-cookbook) Resources for Article: Further resources on this subject: Introduction to Microsoft Azure Cloud Services[article] Microsoft Azure – Developing Web API for Mobile Apps[article] Security in Microsoft Azure[article]
Read more
  • 0
  • 0
  • 19211

Packt
23 Dec 2013
5 min read
Save for later

GLSL – How to Set up the Shaders from the Host Application Side

Packt
23 Dec 2013
5 min read
(For more resources related to this topic, see here.) Setting up geometry Let's say that our mesh (a quad formed by two triangles) has the following information: vertex positions and texture coordinates. Also, we will arrange the data interleaved in the array. struct MyVertex { float x, y, z; float s, t; }; MyVertex geometry[] =  {{0,1,0,0,1}, {0,0,0,0,0}, {1,1,0,1,1},{1,0,0,1,0}}; // Let's create the objects that will encapsulate our geometry data GLuint vaoID, vboID; glGenVertexArray(1, &vaoID); glBindVertexArray(vaoID); glGenBuffers(1, &vboID); glBindBuffer(GL_ARRAY_BUFFER, vboID); // Attach our data to the OpenGL objects glBufferData(GL_ARRAY_BUFFER, 4 * sizeof(MyVertex), &geometry[0].x,GL_DYNAMIC_DRAW); // Specify the format of each vertex attribute glEnableVertexAttribArray(0); glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, sizeof(MyVertex), NULL); glEnableVertexAttribArray(1); glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, sizeof(MyVertex),(void*)(sizeof(float)*3)); At this point, we created the OpenGL objects, set up correctly each vertex attribute format and uploaded the data to GPU. Setting up textures Setting up textures follows the same pattern. First create the OpenGL objects, then fill the data and the format in which it is provided. const int width = 512; const int height = 512; const int bpp = 32; struct RGBColor { unsigned char R,G,B,A; }; RGBColor textureData[width * height]; for(size_t y = 0; y < height; ++y) for(size_t x = 0; x < width; ++x)                textureData[y*height+x] = …; //fill your texture here // Create GL object GLuint texID; glGenTextures(1, &texID); glBindTexture(GL_TEXTURE_2D, texID); // Fill up the data and set the texel format glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0, GL_RGBA,GL_UNSIGNED_BYTE, &textureData[0].R); // Set texture format data: interpolation and clamping modes glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT); glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT); And that's all about textures. Later we will use the texture ID to place the texture into a slot, and that slot number is the information that we will pass to the shader to tell it where the texture is placed in order to locate it. Setting up shaders In order to setup the shaders, we have to carry out some steps:  load up the source code, compile it and associate to a shader object, and link all shaders together into a program object. char* vs[1]; vs[0] = "#version 430\nlayout (location = 0) in vec3 PosIn;layout (location = 1) in vec2 TexCoordIn;smooth out vec2 TexCoordOut;uniform mat4 MVP; void main() { TexCoordOut = TexCoordIn;gl_Position = MVP * vec4(PosIn, 1.0);}"; char fs[1]; fs = "#version 430\n uniform sampler2D Image; smooth in vec2 TexCoord;out vec4 FBColor;void main() {FBColor = texture(Image, TexCoord);}"; // Upload source code and compile it GLuint pId, vsId, fsId; vsId = glCreateShader(GL_VERTEX_SHADER); glShaderSource(vsId, 1, (const char**)&vs, NULL); glCompileShader(vsId); // Check for compilation errors GLint status = 0, bufferLength = 0; glGetShaderiv(vsId, GL_COMPILE_STATUS, &status); if(!status) { char* infolog = new char[bufferLength + 1]; glGetShaderiv(vsId, GL_INFO_LOG_LENGTH, &bufferLength); glGetShaderInfoLog(vsId, bufferLength, NULL, infolog); infolog[bufferLength] = 0; printf("Shader compile errors / warnings: %s\n", infolog); delete [] infolog; } The process for the fragment shader is exactly the same. The only change is that the shader object must be created as fsId = glCreateShader(GL_FRAGMENT_SHADER); // Now let's proceed to link the shaders into the program object pId = glCreateProgram(); glAttachShader(pId, vsId); glAttachShader(pId, fsId); glLinkProgram(pId); glGetProgramiv(pId, GL_LINK_STATUS, &status); if(!status) { char* infolog = new char[bufferLength + 1]; glGetProgramiv(pId, GL_INFO_LOG_LENGTH, &bufferLength); infolog[bufferLength] = 0; printf("Shader linking errors / warnings: %s\n", infolog); delete [] infolog; } // We do not need the vs and fs anymore, so it is same mark them for deletion. glDeleteShader(vsId); glDeleteShader(fsId); The last things to upload to the shader are two uniform variables: the one that corresponds with the view-projection matrix and the one that represents the texture. Those are uniform variables and are set in the following way: // First bind the program object where we want to upload the variables glUseProgram(pId); // Obtain the "slot number" where the uniform is located in GLint location = glGetUniformLocation(pId, "MVP"); float mvp[16] = {1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1}; // Set the data into the uniform's location glUniformMatrix4fv(location, 1, GL_FALSE, mvp); // Active the texture slot 0 glActiveTexture(GL_TEXTURE0); // Bind the texture to the active slot glBindTexture(GL_TEXTURE_2D, texID); location = glGetUniformLocation(pId, "Image"); // Upload the texture's slot number to the uniform variable int imageSlot = 0; glUniform1i(location, imageSlot); And that's all. For the other types of shaders, process is all the same: Create shader object, upload source code, compile and link, but using the proper OpenGL types such as GL_GEOMETRY_SHADER or GL_COMPUTE_SHADER. A last step, to draw all these things, is to establish them as active and issue the draw call: glBindVertexArray(vaoID); glBindTexture(GL_TEXTURE_2D, texID); glUseProgram(pId); glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); Resources for Article: Further resources on this subject: The Basics of GLSL 4.0 Shaders [Article] GLSL 4.0: Using Subroutines to Select Shader Functionality [Article] Getting Started with GLSL [Article]
Read more
  • 0
  • 0
  • 19166

article-image-introduction-using-nodejs-hadoops-mapreduce-jobs
Harri Siirak
25 Sep 2015
5 min read
Save for later

Using Node.js and Hadoop to store distributed data

Harri Siirak
25 Sep 2015
5 min read
Hadoop is a well-known open-source software framework for distributed storage and distributed processing of very large data sets on computer clusters built from commodity hardware. It's designed with a fundamental assumption that hardware failures can (and will) happen and thus should be automatically handled in software by the framework. Under the hood it's using HDFS (Hadoop Distributed File System) for the data storage. HDFS can store large files across multiple machines and it achieves reliability by replicating the data across multiple hosts (default replication factor is 3 and can be configured to be higher when needed). Although it's designed for mostly immutable files and may not be suitable for systems requiring concurrent write-operations. Its target usage is not only restricted to MapReduce jobs, but it also can be used for cost effective and reliable data storage. In the following examples, I am going to give you an overview of how to establish connections to HDFS storage (namenode) and how to perform basic operations on the data. As you can probably guess, I'm using Node.js to build these examples. Node.js is a platform built on Chrome's JavaScript runtime for easily building fast, scalable network applications. Node.js uses an event-driven, non-blocking I/O model that makes it lightweight and efficient, perfect for data-intensive real-time applications that run across distributed devices. So it's really ideal for what I want to show you next. Two popular libraries for acccessing HDFS in Node.js are node-hdfs and webhdfs. The first one uses Hadoop's native libhdfs library and protocol to communicate with Hadoop namenode, albeit it seems to be not maintained anymore and doesn't support Stream API. Another one is using WebHDFS, which defines a public HTTP REST API, directly built into Hadoop's core (namenodes and datanodes both) and which permits clients to access Hadoop from multiple languages without installing Hadoop, and supports all HDFS user operations including reading files, writing to files, making directories, changing permissions and renaming. More details about WebHDFS REST API and about its implementation details and response codes/types can be found from here. At this point I'm assuming that you have Hadoop cluster up and running. There are plenty of good tutorials out there showing how to setup and run Hadoop cluster (single and multi node). Installing and using the webhdfs library webhdfs implements most of the REST API calls, albeit it's not yet supporting Hadoop delegation tokens. It's also Stream API compatible what makes its usage pretty straightforward and easy. Detailed examples and use cases for another supported calls can be found from here. Install webhdfs from npm: npm install wehbhdfs Create a new script named webhdfs-client.js: // Include webhdfs module var WebHDFS = require('webhdfs'); // Create a new var hdfs = WebHDFS.createClient({ user: 'hduser', // Hadoop user host: 'localhost', // Namenode host port: 50070 // Namenode port }); module.exports = hdfs; Here we initialized new webhdfs client with options, including namenode's host and port where we are connecting to. Let's proceed with a more detailed example. Storing file data in HDFS Create a new script named webhdfs-write-test.js and add the code below. // Include created client var hdfs = require('./webhdfs-client'); // Include fs module for local file system operations var fs = require('fs'); // Initialize readable stream from local file // Change this to real path in your file system var localFileStream = fs.createReadStream('/path/to/local/file'); // Initialize writable stream to HDFS target var remoteFileStream = hdfs.createWriteStream('/path/to/remote/file'); // Pipe data to HDFS localFileStream.pipe(remoteFileStream); // Handle errors remoteFileStream.on('error', function onError (err) { // Do something with the error }); // Handle finish event remoteFileStream.on('finish', function onFinish () { // Upload is done }); Basically what we are doing here is that we're initializing readable file stream from a local filesystem and piping its contents seamlessly into remote HDFS target. Optionally webhdfs exposes error and finish. Reading file data from HDFS Let's retrieve the data what we just stored in HDFS storage. Create a new script named webhdfs-read-test.js and add code below. var hdfs = require('./webhdfs-client'); var fs = require('fs'); // Initialize readable stream from HDFS source var remoteFileStream = hdfs.createReadStream('/path/to/remote/file'); // Variable for storing data var data = new Buffer(); remoteFileStream.on('error', function onError (err) { // Do something with the error }); remoteFileStream.on('data', function onChunk (chunk) { // Concat received data chunk data = Buffer.concat([ data, chunk ]); }); remoteFileStream.on('finish', function onFinish () { // Upload is done // Print received data console.log(data.toString()); }); What's next? Now when we have data in Hadoop cluster, we can start processing it by spawning some MapReduce jobs, and when it's processed we can retrieve the output data. In the second part of this article, I'm going to give you an overview of how Node.js can be used as part of MapReduce jobs. About the author Harri is a senior Node.js/Javascript developer among a talented team of full-stack developers who specialize in building scalable and secure Node.js based solutions. He can be found on Github at harrisiirak.
Read more
  • 0
  • 2
  • 19157

article-image-ai-distilled-15-openai-unveils-chatgpt-enterprise-code-llama-by-meta-vulcansql-from-hugging-face-microsofts-algorithm-of-thoughts-google-deepminds-synthid
Merlyn Shelley
31 Aug 2023
14 min read
Save for later

AI_Distilled #15: OpenAI Unveils ChatGPT Enterprise, Code Llama by Meta, VulcanSQL from Hugging Face, Microsoft's "Algorithm of Thoughts”, Google DeepMind's SynthID

Merlyn Shelley
31 Aug 2023
14 min read
👋 Hello ,“[AI] will touch every sector, every industry, every business function, and significantly change the way we live and work..this isn’t just the future. We are already starting to experience the benefits right now. As a company, we’ve been preparing for this moment for some time.” -Sundar Pichai, CEO, Google Speaking at the ongoing Google Cloud Next conference, Pichai emphasized how AI is the future, and it’s here already.   Step into the future with AI_Distilled#15, showcasing the breakthroughs in AI/ML, LLMs, NLP, GPT, and Generative AI, as we talk about Nvidia reporting over 100% increase in sales amid high demand for AI chips, Meta introducing Code Llama: a breakthrough in AI-powered coding assistance, OpenAI introducing ChatGPT Enterprise for businesses, Microsoft’s promising new "Algorithm of Thoughts" to enhance AI reasoning, and Salesforce's State of the Connected Customer Report which shows how businesses are facing AI trust gap with customers. Looking for fresh knowledge resources and tutorials? We’ve got your back! Look out for our curated collection of posts on how to use Code Llama, mitigating hallucination in LLMs, Google’s: Region-Aware Pre-Training for Open-Vocabulary Object Detection with Vision Transformers, and making data queries with Hugging Face's VulcanSQL.  We’ve also handpicked some great GitHub repos for you to use on your next AI project! What do you think of this issue and our newsletter? Please consider taking the short survey below to share your thoughts and you will get a free PDF of the “The Applied Artificial Intelligence Workshop” eBook upon completion. Complete the Survey. Get a Packt eBook for Free!Writer’s Credit: Special shout-out to Vidhu Jain for their valuable contribution to this week’s newsletter content!  Cheers,  Merlyn Shelley  Editor-in-Chief, Packt   ⚡ TechWave: AI/GPT News & Analysis OpenAI Introduces ChatGPT Enterprise: AI Solution for Businesses: OpenAI has unveiled ChatGPT Enterprise with advanced features. The enterprise-grade version offers enhanced security, privacy, and access to the more powerful GPT-4 model. It includes unlimited usage of GPT-4, higher-speed performance, longer context windows for processing lengthier inputs, advanced data analysis capabilities, customization options, and more, targeting improved productivity, customized workflows, and secure data management. Meta Introduces Code Llama: A Breakthrough in AI-Powered Coding Assistance: Code Llama is a cutting-edge LLM designed to generate code based on text prompts and is tailored for code tasks and offers the potential to enhance developer productivity and facilitate coding education. Built on Llama 2, Code Llama comes in different models, including the foundational code model, Python-specialized version, and an instruct variant fine-tuned for understanding natural language instructions. The models outperformed existing LLMs on code tasks and hold promise for revolutionizing coding workflows while adhering to safety and responsible use guidelines. Nvidia Reports Over 100% Increase in Sales Amid High Demand for AI Chips: Nvidia has achieved record-breaking sales, more than doubling its revenue to over $13.5 billion for the quarter ending in June. The company anticipates further growth in the current quarter and plans to initiate a stock buyback of $25 billion. Its stock value soared by more than 6.5% in after-hours trading, bolstering its substantial gains this year. Nvidia's data center business, which includes AI chips, fueled its strong performance, with revenue surpassing $10.3 billion, driven by cloud computing providers and consumer internet firms adopting its advanced processors. With a surge in its market value, Nvidia joined the ranks of trillion-dollar companies alongside Apple, Microsoft, Alphabet, and Amazon. Businesses Facing AI Trust Gap with Customers, Reveals Salesforce's State of the Connected Customer Report: Salesforce's sixth edition of the State of the Connected Customer report highlights a growing concern among businesses about an AI trust gap with their customers. The survey, conducted across 25 countries with over 14,000 consumers and business buyers, indicates that as companies increasingly adopt AI to enhance efficiency and meet customer expectations, nearly three-quarters of their customers are worried about unethical AI use. Consumer receptivity to AI has also decreased over the past year, urging businesses to address this gap by implementing ethical guidelines and providing transparency into AI applications. Microsoft Introduces "Algorithm of Thoughts" to Enhance AI Reasoning: Microsoft has unveiled a novel AI training method called the "Algorithm of Thoughts" (AoT), aimed at enhancing the reasoning abilities of large language models like ChatGPT by combining human-like cognition with algorithmic logic. This new approach leverages "in-context learning" to guide language models through efficient problem-solving paths, resulting in faster and less resource-intensive solutions. The technique outperforms previous methods and can even surpass the algorithm it is based on.  Google's Duet AI Expands Across Google Cloud with Enhanced Features: Google's Duet AI, a suite of generative AI capabilities for tasks like text summarization and data organization, is expanding its reach to various products and services within the Google Cloud ecosystem. The expansion includes assisting with code refactoring, offering guidance on infrastructure configuration and deployment in the Google Cloud Console, writing code in Google's dev environment Cloud Workstations, generating flows in Application Integration, and more. ̌It also integrates generative AI advancements into the security product line. OpenAI Collaborates with Scale to Enhance Enterprise Model Fine-Tuning Support: OpenAI has entered into a partnership with Scale to provide expanded support for enterprises seeking to fine-tune advanced models. Recognizing the demand for high performance and customization in AI deployment, OpenAI introduced fine-tuning for GPT-3.5 Turbo and plans to extend it to GPT-4. This feature empowers companies to customize advanced models with proprietary data, enhancing their utility. OpenAI assures that customer data remains confidential and is not utilized to train other models. Google DeepMind Introduces SynthID: A Tool to Identify AI-Generated Images: In response to the growing prevalence of AI-generated images that can be indistinguishable from real ones, Google Cloud has partnered with Imagen to unveil SynthID. This newly launched beta version aims to watermark and identify AI-created images. The technology seamlessly embeds a digital watermark into the pixels of an image, allowing for imperceptible yet detectable identification. This tool is a step towards responsible use of generative AI and enhances the capacity to identify manipulated or fabricated images.   ✨ Unleashing the Power of Causal Reasoning with LLMs:Join Aleksander Molak on October 11th and be a part of Packt's most awaited event of 2023 on Generative AI! In AI's evolution, a big change is coming. It's all about Causally Aware Prompt Engineering, and you should pay attention because it's important. LLMs are good at recognizing patterns, but what if they could do more? That's where causal reasoning comes in. It's about understanding not just what's connected but why. Let's distill the essence: - LLMs can outperform causal discovery algorithms on some tasks  - GPT-4 achieves a near-human performance on some counterfactual benchmarks  - This might be the case because the models simply memorize the data, but it's also possible that they build a **meta-SCM** (meta structural causal models) based on the correlations of causal facts learned from the data  - LLMs can reason causally if we allow them to intervene on the test time  - LLMs do not reason very well, when we provide them with verbal description of conditional independence structures in the data (but nor do (most of) humans). Now, catalyze your journey with three simple techniques: Causal Effect Estimation: Causal effect estimate aims at capturing the strength of (expected) change in the outcome variable when we modify the value of the treatment by one unit. In practice, almost any machine learning algorithm can be used for this purpose, yet in most cases we need to use these algorithms in a way that differs from the classical machine learning flow. Confronting Confounding: The main challenge (yet not the only one) in estimating causal effects from observational data comes from confounding. Confounder is a variable in the system of interest that produces a spurious relationship between the treatment and the outcome. Spurious relationships are a kind of illusion. Interestingly, you can observe spurious relationships not only in the recorded data, but also in the real world. Unveiling De-confounding: To obtain an unbiased estimate of the causal effect, we need to get rid of confounding. At the same time, we need to be careful not to introduce confounding ourselves! This usually boils down to controlling for the right subset of variables in your analysis. Not too small, not too large. If you're intrigued by this, I invite you to join me for an in-depth exploration of this fascinating topic at Packt's upcoming Generative AI conference on October 11th. During my power-talk, we'll delve into the question: Can LLMs learn Causally?  REGISTER NOW at Early Bird discounted pricing! *Free eBook on Registration: Modern Generative AI with ChatGPT and OpenAI Models   🔮 Expert Insights from Packt Community The Regularization Cookbook - By Vincent Vandenbussche Regularization serves as a valuable approach to enhance the success rate of ML models in production. Effective regularization techniques can prevent AI recruitment models from exhibiting gender biases, either by eliminating certain features or incorporating synthetic data. Additionally, proper regularization enables chatbots to maintain an appropriate level of sensitivity toward new tweets. It also equips models to handle edge cases and previously unseen data proficiently, even when trained on synthetic data. Key concepts of regularization Let us now delve into a more precise definition and explore key concepts that enable us to better comprehend regularization. Bias and variance Bias and variance are two key concepts when talking about regularization. We can define two main kinds of errors a model can have: Bias is how bad a model is at capturing the general behavior of the data Variance is how bad a model is at being robust to small input data fluctuations Let’s describe those four cases: High bias and low variance: The model is hitting away from the center of the target, but in a very consistent manner Low bias and high variance: The model is, on average, hitting the center of the target, but is quite noisy and inconsistent in doing so High bias and high variance: The model is hitting away from the center in a noisy way Low bias and low variance: The best of both worlds – the model is hitting the center of the target consistently  The above content is extracted from the book The Regularization Cookbook By Vincent Vandenbussche and published in July 2023. To get a glimpse of the book's contents, make sure to read the free chapter provided here, or if you want to unlock the full Packt digital library free for 7 days, try signing up now! To learn more, click on the button below. Keep Calm, Start Reading!  🌟 Secret Knowledge: AI/LLM Resources Google’s RO-ViT: Region-Aware Pre-Training for Open-Vocabulary Object Detection with Vision Transformers: Google's research scientists have unveiled a new method called "RO-ViT" that enhances open-vocabulary object detection using vision transformers. Learn how the technique addresses limitations in existing pre-training approaches for vision transformers, which struggle to fully leverage the concept of objects or regions during pre-training. RO-ViT introduces a novel approach called "cropped positional embedding" that aligns better with region-level tasks.Tiered AIOps: Enhancing Cloud Platform Management with AI: Explore the concept of Tiered AIOps to manage complex cloud platforms. The ever-changing nature of cloud applications and infrastructure presents challenges for complete automation, requiring a tiered approach to combine AI and human intervention. The concept involves dividing operations into tiers, each with varying levels of automation and human expertise. Tier 1 incorporates routine operations automated by AI, Tier 2 empowers non-expert operators with AI assistance, and Tier 3 engages expert engineers for complex incidents. Effective AI-Agent Interaction: SERVICE Principles Unveiled: In this post, you'll learn how to design AI agents that can interact seamlessly and effectively with users, aiming to transition from self-service to "agent-service." The author introduces the concept of autonomous AI agents capable of performing tasks on users' behalf and offers insights into their potential applications. The SERVICE principles, rooted in customer service and hospitality practices, are presented as guidelines for designing agent-user interactions. These principles encompass key aspects like salient responses, explanatory context, reviewable inputs, vaulted information, indicative guidance, customization, and empathy.  How to Mitigate Hallucination in Large Language Models: In this article, researchers delve into the persistent challenge of hallucination in Generative LLMs. The piece explores the reasons behind LLMs generating nonsensical or non-factual responses, and the potential consequences for system reliability. The focus is on practical approaches to mitigate hallucination, including adjusting the temperature parameter, employing thoughtful prompt engineering, and incorporating external knowledge sources. The authors conduct experiments to evaluate different methods, such as Chain of Thoughts, Self-Consistency, and Tagged Context Prompts.    💡 MasterClass: AI/LLM Tutorials How to Use Code Llama: A Breakdown of Features and Usage: Code Llama has made a significant stride in code-related tasks, offering an open-access suite of models specialized for code-related challenges. This release includes various notable components, such as integration within the Hugging Face ecosystem, transformative integration, text generation inference, and inference endpoints. Learn how these models showcase remarkable performance across programming languages, enabling enhanced code understanding, completion, and infilling.  Make Data Queries with Hugging Face's VulcanSQL: In this post, you'll learn how to utilize VulcanSQL, an open-source data API framework, to streamline data queries. VulcanSQL integrates Hugging Face's powerful inference capabilities, allowing data professionals to swiftly generate and share data APIs without extensive backend knowledge. By incorporating Hugging Face's Inference API, VulcanSQL enhances the efficiency of query processes. The framework's HuggingFace Table Question Answering Filter offers a unique solution by leveraging pre-trained AI models for NLP tasks.  Exploring Metaflow and Ray Integration for Supercharged ML Workflows: Explore the integration of Metaflow, an extensible ML orchestration framework, with Ray, a distributed computing framework. This collaboration leverages AWS Batch and Ray for distributed computing, enhancing Metaflow’s capabilities. Know how this integration empowers Metaflow users to harness Ray’s features within their workflows. The article also delves into the challenges faced, the technical aspects of the integration, and real-world test cases, offering valuable insights into building efficient ML workflows using these frameworks. Explore Reinforcement Learning Through Solving Leetcode Problems: Explore how reinforcement learning principles can be practically grasped by solving a Leetcode problem. The article centers around the "Shortest Path in a Grid with Obstacles Elimination" problem, where an agent aims to find the shortest path from a starting point to a target in a grid with obstacles, considering the option to eliminate a limited number of obstacles. Explore the foundations of reinforcement learning, breaking down terms like agent, environment, state, and reward system. The author provides code examples and outlines how a Q-function is updated through iterations.    🚀 HackHub: Trending AI Tools apple/ml-fastvit: Introduces a rapid hybrid ViT empowered by structural reparameterization for efficient vision tasks. openchatai/opencopilot: A personal AI copilot repository that seamlessly integrates with APIs and autonomously executes API calls using LLMs, streamlining developer tasks and enhancing efficiency. neuml/txtai: An embeddings database for advanced semantic search, LLM orchestration, and language model workflows featuring vector search, multimodal indexing, and flexible pipelines for text, audio, images, and more. Databingo/aih: Interact with AI models via terminal (Bard, ChatGPT, Claude2, and Llama2) to explore diverse AI capabilities directly from your command line. osvai/kernelwarehouse: Optimizes dynamic convolution by redefining kernel concepts, improving parameter dependencies, and increasing convolutional efficiency. morph-labs/rift: Open-source AI-native infrastructure for IDEs, enabling collaborative AI software engineering. mr-gpt/deepeval: Python-based solution for offline evaluations of LLM pipelines, simplifying the transition to production. 
Read more
  • 0
  • 0
  • 19155

article-image-asynchronous-control-flow-patterns-es2015-and-beyond
Packt
07 Jun 2016
6 min read
Save for later

Asynchronous Control Flow Patterns with ES2015 and beyond

Packt
07 Jun 2016
6 min read
In this article,by Luciano Mammino, the author of the book Node.js Design Patterns, Second Edition, we will explore async await, an innovative syntaxthat will be available in JavaScript as part of the release of ECMAScript 2017. (For more resources related to this topic, see here.) Async await using Babel Callbacks, promises, and generators turn out to be the weapons at our disposal to deal with asynchronous code in JavaScript and in Node.js. As we have seen, generators are very interesting because they offer a way to actually suspend the execution of a function and resume it at a later stage. Now we can adopt this feature to write asynchronous codethatallowsdevelopers to write functions that "appear" to block at each asynchronous operation, waiting for the results before continuing with the following statement. The problem is that generator functions are designed to deal mostly with iterators and their usage with asynchronous code feels a bit cumbersome.It might be hard to understand,leading to code that is hard to read and maintain. But there is hope that there will be a cleaner syntax sometime in the near future. In fact, there is an interesting proposal that will be introduced with the ECMAScript 2017 specification that defines the async function's syntax. You can read more about the current status of the async await proposal at https://tc39.github.io/ecmascript-asyncawait/. The async function specification aims to dramatically improve the language-level model for writing asynchronous code by introducing two new keywords into the language: async and await. To clarify how these keywords are meant to be used and why they are useful, let's see a very quick example: const request = require('request'); function getPageHtml(url) { return new Promise(function(resolve, reject) { request(url, function(error, response, body) { resolve(body); }); }); } async function main() { const html = awaitgetPageHtml('http://google.com'); console.log(html); } main(); console.log('Loading...'); In this code,there are two functions: getPageHtml and main. The first one is a very simple function that fetches the HTML code of a remote web page given its URL. It's worth noticing that this function returns a promise. The main function is the most interesting one because it's where the new async and await keywords are used. The first thing to notice is that the function is prefixed with the async keyword. This means that the function executes asynchronous code and allows it to use the await keyword within its body. The await keyword before the call to getPageHtml tells the JavaScript interpreter to "await" the resolution of the promise returned by getPageHtml before continuing to the next instruction. This way, the main function is internally suspended until the asynchronous code completes without blocking the normal execution of the rest of the program. In fact, we will see the string Loading… in the console and, after a moment, the HTML code of the Google landing page. Isn't this approach much more readable and easy to understand? Unfortunately, this proposal is not yet final, and even if it will be approved we will need to wait for the next version of the ECMAScript specification to come out and be integrated in Node.js to be able to use this new syntax natively. So what do we do today? Just wait? No, of course not! We can already leverage async await in our code thanks to transpilers such as Babel. Installing and running Babel Babel is a JavaScript compiler (or transpiler) that is able to convert JavaScript code into other JavaScript code using syntax transformers. Syntax transformers allowsthe use of new syntax such as ES2015, ES2016, JSX, and others to produce backward compatible equivalent code that can be executed in modernJavaScript runtimes, such as browsers or Node.js. You can install Babel in your project using NPM with the following command: npm install --save-dev babel-cli We also need to install the extensions to support async await parsing and transformation: npm install --save-dev babel-plugin-syntax-async-functions babel-plugin-transform-async-to-generator Now let's assume we want to run our previous example (called index.js).We need to launch the following command: node_modules/.bin/babel-node --plugins "syntax-async-functions,transform-async-to-generator" index.js This way, we are transforming the source code in index.js on the fly, applying the transformers to support async await. This new backward compatible code is stored in memory and then executed on the fly on the Node.js runtime. Babel can also be configured to act as a build processor that stores the generated code into files so that you can easily deploy and run the generated code. You can read more about how to install and configure Babel on the official website at https://babeljs.io. Comparison At this point, we should have a better understanding of the options we have to tame the asynchronous nature of JavaScript. Each one of the solutions presented has its own pros and cons. Let's summarize them in the following table: Solutions Pros Cons Plain JavaScript Does not require any additional libraries or technology Offers the best performances Provides the best level of compatibility with third-party libraries Allows the creation of ad hoc and more advanced algorithms Might require extra code and relatively complex algorithms Async (library) Simplifies the most common control flow patterns Is still a callback-based solution Good performance Introduces an external dependency Might still not be enough for advanced flows   Promises Greatly simplify the most common control flow patterns Robust error handling Part of the ES2015 specification Guarantee deferred invocation of onFulfilled and onRejected Require to promisify callback-based APIs Introduce a small performance hit   Generators Make non-blocking API look like a blocking one Simplify error handling Part of ES2015 specification Require a complementary control flow library Still require callbacks or promises to implement non-sequential flows Require to thunkify or promisify nongenerator-based APIs   Async await Make non-blocking API look like blocking Clean and intuitive syntax Not yet available in JavaScript and Node.js natively Requires Babel or other transpilers and some configuration to be used today   It is worth mentioning that we chose to present only the most popular solutions to handle asynchronous control flow, or the ones receiving a lot of momentum, but it's good to know that there are a few more options you might want to look at, for example, Fibers (https://npmjs.org/package/fibers) and Streamline (https://npmjs.org/package/streamline). Summary In this article, we analyzed how Babel can be used for performing async await and how to install Babel.
Read more
  • 0
  • 0
  • 19131

article-image-openstack-networking-nutshell
Packt
22 Sep 2016
13 min read
Save for later

OpenStack Networking in a Nutshell

Packt
22 Sep 2016
13 min read
Information technology (IT) applications are rapidly moving from dedicated infrastructure to cloud based infrastructure. This move to cloud started with server virtualization where a hardware server ran as a virtual machine on a hypervisor. The adoptionof cloud based applicationshas accelerated due to factors such as globalization and outsourcing where diverse teams need to collaborate in real time. Server hardware connects to network switches using Ethernet and IP to establish network connectivity. However, as servers move from physical to virtual, the network boundary also moves from the physical network to the virtual network.Traditionally applications, servers and networking were tightly integrated. But modern enterprises and IT infrastructure demand flexibility in order to support complex applications. The flexibility of cloud infrastructure requires networking to be dynamic and scalable. Software Defined Networking (SDN) and Network Functions Virtualization (NFV) play a critical role in data centers in order to deliver the flexibility and agility demanded by cloud based applications. By providing practical management tools and abstractions that hide underlying physical network’s complexity, SDN allows operators to build complex networking capabilities on demand. OpenStack is an open source cloud platform that helps build public and private cloud at scale. Within OpenStack, the name for OpenStack Networking project is Neutron. The functionality of Neutron can be classified as core and service. In this article by Sriram Subramanian and SreenivasVoruganti, authors of the book Software Defined Networking (SDN) with OpenStack, aims to provide a short introduction toOpenStack Networking. We will cover the following topics in this article: Understand traffic flows between virtual and physical networks Neutron entities that support Layer 2 (L2) networking Layer 3 (L3) or routing between OpenStack Networks Securing OpenStack network traffic Advanced networking services in OpenStack OpenStack and SDN The terms Neutron and OpenStack Networking are used interchangeably throughout this article. (For more resources related to this topic, see here.) Virtual and physical networking Server virtualization led to the adoption of virtualized applications and workloads running inside physical servers. While physical servers are connected to the physical network equipment, modern networking has pushed the boundary of networking into the virtual domain as well. Virtual switches, firewalls and routers play a critical role in the flexibility provided by cloud infrastructure. Figure 1: Networking components for server virtualization The preceding figure describes a typical virtualized server and the various networking components. The virtual machines are connected to a virtual switch inside the compute node (or server). The traffic is secured using virtual routers and firewalls. The compute node is connected to a physical switch which is the entry point into the physical network. Let us now walk through different traffic flow scenarios using the picture above as the background. In Figure 2, traffic from one VM to another on same compute node is forwarded by the virtual switch itself. It does not reach the physical network. You can even apply firewall rules to traffic between the two virtual machine. Figure 2: Traffic flow between two virtual machines on the same server Next, let us have a look at how traffic flows between virtual machines across two compute nodes. In Figure 3, the traffic comes out from compute node and then reaches the physical switch. The physical switch forwards the traffic to the second compute node and the virtual switch within the second compute node steers the traffic to appropriate VM. Figure 3: Traffic flow between two virtual machines on the different servers Finally, here is the depiction of traffic flow when a virtual machine sends or receives traffic from the Internet. The physical switch forwards the traffic to the physical router and firewall which is presumed to be connected to the internet. Figure 4: Traffic flow from a virtual machine to external network As seen from the above diagrams, the physical and the virtual network components work together to provide connectivity to virtual machines and applications. Tenant isolation As a cloud platform, OpenStack supports multiple users grouped into tenants. One of the key requirements of a multi-tenant cloud is to provide isolation of data traffic belonging to one tenant from rest of the tenants that use the same infrastructure. OpenStack supports different ways of achieving isolation and the it is the responsibility of the virtual switch to implement the isolation. Layer 2 (L2) capabilities in OpenStack The connectivity to a physical or virtual switch is also known as Layer 2 (L2) connectivity in networking terminology. Layer 2 connectivity is the most fundamental form of network connectivity needed for virtual machines. As mentioned earlier OpenStack supports core and service functionality. The L2 connectivity for virtual machines falls under the core capability of OpenStack Networking, whereas Router, Firewall etc., fall under the service category. The L2 connectivity in OpenStack is realized using two constructs called Network and Subnet. Operators can use OpenStack CLI or the web interface to create Networks and Subnets. And virtual machines are instantiated, the operators can associate them appropriate Networks. Creatingnetwork using OpenStack CLI A Network defines the Layer 2 (L2) boundary for all the instances that are associated with it. All the virtual machines within a Network are part of the same L2 broadcast domain. The Liberty release has introduced new OpenStack CLI (Command Line Interface) for different services. We will use the new CLI and see how to create a Network. Creating Subnet using OpenStack CLI A Subnet is a range of IP addresses that are assigned to virtual machines on the associated network. OpenStack Neutron configures a DHCP server with this IP address range and it starts one DHCP server instance per Network, by default. We will now show you how to create a Subnet using OpenStack CLI. Note: Unlike Network, for Subnet, we need to use the regular neutron CLI command in the Liberty release. Associating a network and Subnet to a virtual machine To give a complete perspective, we will create a virtual machine using OpenStack web interface and show you how to associate a Network and Subnet to a virtual machine. In your OpenStack web interface, navigate to Project|Compute|Instances. Click on Launch Instances action on the right hand side as highlighted above. In the resulting window enter the name for your instance and how you want to boot your instance. To associate a network and a subnet with the instance, click on Networking tab. If you have more than one tenant network, you will be able to choose the network you want to associate with the instance. If you have exactly one network, the web interface will automatically select it. As mentioned earlier, providing isolation for Tenant network traffic is a key requirement for any cloud. OpenStack Neutron uses Network and Subnet to define the boundaries and isolate data traffic between different tenants. Depending on Neutron configuration, the actual isolation of traffic is accomplished by the virtual switches. VLAN and VXLAN are common networking technologies used to isolate traffic. Layer 3 (L3) capabilities in OpenStack Once L2 connectivity is established, the virtual machines within one Network can send or receive traffic between themselves. However, two virtual machines belonging to two different Networks will not be able communicate with each other automatically. This is done to provide privacy and isolation for Tenant networks. In order to allow traffic from one Network to reach another Network, OpenStack Networking supports an entity called Router. The default implementation of OpenStack uses Namespaces to support L3 routing capabilities. Creating Router using OpenStack CLI Operators can create Routers using OpenStack CLI or web interface. They can then add more than one Subnets as interface to the Router. This allows the Networks associated with the router to exchange traffic with one another. The command to create a Router is as follows: This command creates a Router with the specified name. Associating Subnetwork to a Router Once a Router is created, the next step is to associate one or more sub-networks to the Router. The command to accomplish this is: The Subnet represented by subnet1 is now associated to the Router router1. Securing network traffic in OpenStack The security of network traffic is very critical and OpenStack supports two mechanisms to secure network traffic. Security Groups allow traffic within a tenant’s network to be secured. Linux iptables on the compute nodes are used to implement OpenStack security groups. The traffic that goes outside of a tenant’s network – to another Network or the Internet, is secured using OpenStackFirewall Service functionality. Like Routing, Firewall is a service with Neutron. Firewall service also uses iptables but the scope of iptables is limited to the OpenStack Router used as part of the Firewall Service. Usingsecurity groups to secure traffic within a network In order to secure traffic going from one VM to another within a given Network, we must create a security group. The command to create a security group is: The next step is to create one or more rules within the security group. As an example let us create a rule which allows only UDP, incoming traffic on port 8080 from any Source IP address. The final step is to associate this security group and the rules to a virtual machine instance. We will use the nova boot command for this: Once the virtual machine instance has a security group associated with it, the incoming traffic will be monitored and depending upon the rules inside the security group, data traffic may be blocked or permitted to reach the virtual machine. Note: it is possible to block ingress or egress traffic using security groups. Using firewall service to secure traffic We have seen that security groups provide a fine grain control over what traffic is allowed to and from a virtual machine instance. Another layer of security supported by OpenStack is the Firewall as a Service (FWaaS). The FWaaS enforces security at the Router level whereas security groups enforce security at a virtual machine interface level. The main use case of FWaaS is to protect all virtual machine instances within a Network from threats and attacks from outsidethe Network. This could be virtual machines part of another Network in the same OpenStack cloud or some entity on the Internet trying to make an unauthorized access. Let us now see how FWaaS is used in OpenStack. In FWaaS, a set of firewall rules are grouped into a firewall policy. And then a firewall is created that implements one policy at a time. This firewall is then associated to a Router. Firewall rule can be created using neutron firewall-rule-create command as follows: This rule blocks ICMP protocol so applications like Ping will be blocked by the firewall. The next step is to create a Firewall policy. In real world scenarios the security administrators will define several rules and consolidate them under a single policy. For example all rules that block various types of traffic can be combined into a single Policy. The command to create a firewall policy is: The final step is to create a firewall and associate it with a router. The command to do this is: In the command above we did not specify any Routers and the OpenStack behavior is to associate the firewall (and in turn the policy and rules) to all the Routers available for that tenant. The neutron firewall-create command supports an option to pick a specific Router as well. Advanced networking services Besides routing and firewall, there are few other commonly used networking technologies supported by OpenStack. Let’s take a quick look at these without delving deep into the respective commands. Load Balancing as a Service (LBaaS) Virtual machines instances created in OpenStack are used to run applications. Most applications are required to support redundancy and concurrent access. For example, a web server may be accessed by a large number of users at the same time. One of the common strategies to handle scale and redundancy is to implement load-balancing for incoming requests. In this approach, aLoad Balancer distributes an incoming service request onto a pool of servers, which processes the request thus providing higher throughput. If one of the servers in the pool fails, the Load Balancer removes it from the pool and the subsequent service requests are distributed among the remaining servers. User of the application use the IP address of the Load Balancer to access the application and are unaware of the pool of servers. OpenStack implements Load Balancer using HAProxy software and Linux Namespace. Virtual Private Network as a Service (VPNaaS) As mentioned earlier tenant isolation requires data traffic to be segregated and secured within an OpenStack cloud. However, there are times when external entities need to be part of the same Network without removing the firewall based security. This can be accomplished using a Virtual Private Network or VPN. A Virtual Private Network (VPN) connects two endpoints on different networks over a public Internet connection, such that the endpoints appear to be directly connected to each other.  VPNs also provide confidentiality and integrity of transmitted data. Neutron provides a service plugin that enables OpenStack users to connect two networks using a Virtual Private Network (VPN).  The reference implementation of VPN plugin in Neutron uses Openswan to create an IPSec based VPN. IPSec is a suite of protocols that provides secure connection between two endpoints by encrypting each IP packet transferred between them. OpenStack and SDN context So far in this article we have seen the different networking capabilities provided by OpenStack. Let us know look at two capabilities in OpenStack that enable SDN to be leveraged effectively. Choice of technology OpenStack being an open source platform bundles open source networking solutions as default implementation for these networking capabilities. For example, Routing is supported using Namespace, security using iptables and Load balancing using HAproxy. Historically these networking capabilities were implemented using customized hardware and software, most of them being proprietary solutions. These custom solutions are capable of much higher performance and are well supported by their vendors. And hence they have a place in the OpenStack and SDN ecosystem. From it initial releases OpenStack has been designed for extensibility. Vendors can write their own extensions and then can easily configure OpenStack to use their extension instead of the default solutions. This allows the operators to deploy the networking technology of their choice. OpenStack API for networking One of the most powerful capabilities of OpenStack is the extensive support for APIs. All services of OpenStack interact with one another using well defined RESTful APIs. This allows custom implementations and pluggable components to provide powerful enhancements for practical cloud implementation. For example, when a Network is created using OpenStack web interface, a RESTful request is sent to Horizon service. This in turn invokes a RESTful API to validate the user using Keystone service. Once validated, Horizon sends another RESTful API request to Neutron to actually create the Network. Summary As seen in this article, OpenStack supports a wide variety of networking functionality right out of the box. The importance of isolating tenant traffic and the need to allow customized solution requires OpenStack to support flexible configuration. We also highlighted some key aspects of OpenStack that will play a key role in deploying Software Defined Networking in datacenters, thereby supporting powerful cloud architecture and solution. Resources for Article: Further resources on this subject: Setting Up a Network Backup Server with Bacula [article] Jenkins 2.0: The impetus for DevOps Movement [article] Integrating Accumulo into Various Cloud Platforms [article]
Read more
  • 0
  • 0
  • 19085
Unlock access to the largest independent learning library in Tech for FREE!
Get unlimited access to 7500+ expert-authored eBooks and video courses covering every tech area you can think of.
Renews at $19.99/month. Cancel anytime
article-image-magic-leaps-first-mixed-reality-headset-powered-by-nvidia-tegra-x2-is-coming-this-summer
Richard Gall
16 Jul 2018
2 min read
Save for later

Magic Leap's first augmented reality headset, powered by Nvidia Tegra X2, is coming this Summer

Richard Gall
16 Jul 2018
2 min read
The Magic Leap One - the first augmented reality headset produced by startup Magic Leap - is slated to be launched this Summer. The news closely follows an announcement by AT&T that it will be the sole carrier for the headset. However, although Magic Leap revealed a lot about how the Magic Leap One in a live stream on Friday (13 July), no official release date has been stated. What we learned from the Magic Leap One livestream The Magic Leap One livestream gave us a pretty neat insight into how the virtual reality headset is going to work. It showed us how gestures form the main part of the UX with a demo of a rock-throwing game called 'Dodge'. The visuals looked pretty exciting, and it seems like Magic Leap have done a good job of developing a headset that could be game-changing for the augmented reality industry. However, there were a few holes - one Twitter user noticed, for example, that the software at one point failed to recognize when a user's hand would have been blocking the virtual reality images. https://twitter.com/ID_R_McGregor/status/1017119982906494983 For those with a particular interest in the engineering that has gone into the headset, we also found out that the Magic Leap One runs on an Nvidia Tegra X2 processor. This makes it considerably more powerful than the processor that is helping to power the current Nintendo Switch console, which uses the Tegra X1.
Read more
  • 0
  • 0
  • 19085

article-image-using-tiled-map-editor
Packt
13 Oct 2015
5 min read
Save for later

Using the Tiled map editor

Packt
13 Oct 2015
5 min read
LibGDX is a game framework and not a game engine. This is why it doesn't have any editor to place the game objects or to make levels. Tiled is a 2D level/map editor well-suited for this purpose. In this article by Indraneel Potnis, the author of LibGDX Cross-platform Development Blueprints, we will learn how to draw objects and create animations. LibGDX has an excellent support for rendering and reading maps/levels made through Tiled. (For more resources related to this topic, see here.) Drawing objects Sometimes, simple tiles may not satisfy your requirements. You might need to create objects with complex shapes. You can define these shape outlines in the editor easily. The first thing you need to do is create an object layer. Go to Layer | Add Object Layer: You will notice that a new layer has been added to the Layers pane called Object Layer 1. You can rename it if you like: With this layer selected, you can see the object toolbar getting enabled: You can draw basic shapes, such as a rectangle or an ellipse/circle: You can also draw a polygon and a polyline by selecting the appropriate options from the toolbar. Once you have added all the edges, click on the right mouse button to stop drawing the current object: Once the polygon/polyline is drawn, you can edit it by selecting the Edit Polygons option from the toolbar: After this, select the area that encompasses your polygon in order to change to the edit mode. You can edit your polygons/polylines now: You can also add custom properties to your polygons by right-clicking on them and selecting Object Properties: You can then add custom properties as mentioned previously: You can also add tiles as an object. Click on the Insert Tile icon in the toolbar: Once you select this, you can insert tiles as objects into the map. You will observe that the tiles can be placed anywhere now, irrespective of the grid boundaries: To select and move multiple objects, you can select the Select Objects option from the toolbar: You can then select the area that encompasses the objects. Once they are selected, you can move them by dragging them with your mouse cursor: You can also rotate the object by dragging the indicators at the corners after they are selected: Tile animations and images Tiled allows you to create animations in the editor. Let's make an animated shining crystal. First, we will need an animation sheet of the crystal. I am using this one, which is 16 x 16 pixels per crystal: The next thing we need to do is add this sheet as a tileset to the editor and name it crystals. After you add the tileset, you can see a new tab in the Tilesets pane: Go to View | Tile Animation Editor to open the animation editor: A new window will open that will allow you to edit the animations: On the right-hand side, you will see the individual animation frames that make up the animation. This is the animation tileset, which we added. Hold Ctrl on your keyboard, and select all of them with your mouse. Then, drag them to the left window: The numbers beside the images indicate the amount of time each image will be displayed in milliseconds. The images are displayed in this order and repeat continuously. In this example, every image will be shown for 100ms or 1/10th of a second. In the bottom-left corner, you can preview the animation you just created. Click on the Close button. You can now see something like this in the Tilesets pane: The first tile represents the animation, which we just created. Select it, and you can draw the animation anywhere in the map. You can see the animation playing within the map: Lastly, we can also add images to our map. To use them, we need to add an image layer to our map. Go to Layer | Add Image Layer. You will notice that a new layer has been added to the Layers pane. Rename it House: To use an image, we need to set the image's path as a property for this layer. In the Properties pane, you will find a property called Image. There is a file picker next to it where you can select the image you want: Once you set the image, you can use it to draw on the map:   Summary In this article, we learned about a tool called Tiled, and we also learned how to draw various objects and make tile animations and add images. Carry on with LibGDX Cross-platform Development Blueprints to learn how to develop great games, such as Monty Hall Simulation, Whack a Mole, Bounce the Ball, and many more. You can also take a look at the vast array of LibGDX titles from Packt Publishing, a few among these are as follows: Learning Libgdx Game Development, Andreas Oehlke LibGDX Game Development By Example, James Cook LibGDX Game Development Essentials, Juwal Bose Resources for Article:   Further resources on this subject: Getting to Know LibGDX [article] Using Google's offerings [article] Animations in Cocos2d-x [article]
Read more
  • 0
  • 0
  • 19074

article-image-building-games-html5-and-dart
Packt
21 Sep 2015
19 min read
Save for later

Building Games with HTML5 and Dart

Packt
21 Sep 2015
19 min read
In this article written by Ivo Balbaert, author of the book Learning Dart - Second Edition, you will learn to create a well-known memory game. Also, you will design a model first and work up your way from a modest beginning to a completely functional game, step by step. You will also learn how to enhance the attractiveness of web games with audio and video techniques. The following topics will be covered in this article: The model for the memory game Spiral 1—drawing the board Spiral 2—drawing cells Spiral 3—coloring the cells Spiral 4—implementing the rules Spiral 5—game logic (bringing in the time element) Spiral 6—some finishing touches Spiral 7—using images (For more resources related to this topic, see here.) The model for the memory game When started, the game presents a board with square cells. Every cell hides an image that can be seen by clicking on the cell, but this disappears quickly. You must remember where the images are, because they come in pairs. If you quickly click on two cells that hide the same picture, the cells will "flip over" and the pictures will stay visible. The objective of the game is to turn over all the pairs of matching images in a very short time. After some thinking we came up with the following model, which describes the data handled by the application. In our game, we have a number of pictures, which could belong to a Catalog. For example, a travel catalog with a collection of photos from our trips or something similar. Furthermore, we have a collection of cells and each cell is hiding a picture. Also, we have a structure that we will call memory, and this contains the cells in a grid of rows and columns. We could draw it up as shown in the following figure. You can import the model from the game_memory_json.txt file that contains its JSON representation: A conceptual model of the memory game The Catalog ID is its name, which is mandatory, but the description is optional. The Picture ID consists of the sequence number within the Catalog. The imageUri field stores the location of the image file. width and height are optional properties, since they may be derived from the image file. The size may be small, medium, or large to help select an image. The ID of a Memory is its name within the Catalog, the collection of cells is determined by the memory length, for example, 4 cells per side. Each cell is of the same length cellLength, which is a property of the memory. A memory is recalled when all the image pairs are discovered. Some statistics must be kept, such as recall count, the best recall time in seconds, and the number of cell clicks to recover the whole image (minTryCount). The Cell has the row and column coordinates and also the coordinates of its twin with the same image. Once the model is discussed and improved, model views may be created: a Board would be a view of the Memory concept and a Box would be a view of the Cell concept. The application would be based on the Catalog concept. If there is no need to browse photos of a catalog and display them within a page, there would not be a corresponding view. Now, we can start developing this game from scratch. Spiral 1 – drawing the board The app starts with main() in educ_memory_game.dart: library memory; import 'dart:html'; part 'board.dart'; void main() { // Get a reference to the canvas. CanvasElement canvas = querySelector('#canvas'); (1) new Board(canvas); (2) } We'll draw a board on a canvas element. So, we need a reference that is given in line (1). The Board view is represented in code as its own Board class in the board.dart file. Since everything happens on this board, we construct its object with canvas as an argument (line (2)). Our game board will be periodically drawn as a rectangle in line (4) by using the animationFrame method from the Window class in line (3): part of memory; class Board { CanvasElement canvas; CanvasRenderingContext2D context; num width, height; Board(this.canvas) { context = canvas.getContext('2d'); width = canvas.width; height = canvas.height; window.animationFrame.then(gameLoop); (3) } void gameLoop(num delta) { draw(); window.animationFrame.then(gameLoop); } void draw() { clear(); border(); } void clear() { context.clearRect(0, 0, width, height); } void border() { context..rect(0, 0, width, height)..stroke(); (4) } } This is our first result: The game board Spiral 2 – drawing cells In this spiral, we will give our app code some structure: Board is a view, so board.dart is moved to the view folder. We will also introduce here the Memory class from our model in its own code memory.dart file in the model folder. So, we will have to change the part statements to the following: part 'model/memory.dart'; part 'view/board.dart'; The Board view needs to know about Memory. So, we will include it in the Board class and make its object in the Board constructor: new Board(canvas, new Memory(4)); The Memory class is still very rudimentary with only its length property: class Memory { num length; Memory(this.length); } Our Board class now also needs a method to draw the lines, which we decided to make private because it is specific to Board, as well as the clear() and border()methods: void draw() { _clear(); _border(); _lines(); } The lines method is quite straightforward; first draw it on a piece of paper and translate it to code using moveTo and lineTo. Remember that x goes from top-left to right and y goes from top-left to bottom: void _lines() { var gap = height / memory.length; var x, y; for (var i = 1; i < memory.length; i++) { x = gap * i; y = x; context ..moveTo(x, 0) ..lineTo(x, height) ..moveTo(0, y) ..lineTo(width, y); } } The result is a nice grid: Board with cells Spiral 3 – coloring the cells To simplify, we will start using colors instead of pictures to be shown in the grid. Up until now, we didn't implement the cell from the model. Let's do that in modelcell.dart. We start simple by saying that the Cell class has the row, column, and color properties, and it belongs to a Memory object passed in its constructor: class Cell { int row, column; String color; Memory memory; Cell(this.memory, this.row, this.column); } Because we need a collection of cells, it is a good idea to make a Cells class, which contains List. We give it an add method and also an iterator so that we are able to use a for…in statement to loop over the collection: class Cells { List _list; Cells() { _list = new List(); } void add(Cell cell) { _list.add(cell); } Iterator get iterator => _list.iterator; } We will need colors that are randomly assigned to the cells. We will also need some utility variables and methods that do not specifically belong to the model and don't need a class. Hence, we will code them in a folder called util. To specify the colors for the cells, we will use two utility variables: a List variable of colors (colorList), which has the name colors, and a colorMap variable that maps the names to their RGB values. Refer to utilcolor.dart; later on, we can choose some fancier colors: var colorList = ['black', 'blue', //other colors ]; var colorMap = {'black': '#000000', 'blue': '#0000ff', //... }; To generate (pseudo) random values (ints, doubles, or Booleans), Dart has the Random class from dart:math. We will use the nextInt method, which takes an integer (the maximum value) and returns a positive random integer in the range from 0 (inclusive) to max (exclusive). We will build upon this in utilrandom.dart to make methods that give us a random color: int randomInt(int max) => new Random().nextInt(max); randomListElement(List list) => list[randomInt(list.length - 1)]; String randomColor() => randomListElement(colorList); String randomColorCode() => colorMap[randomColor()]; Our Memory class now contains an instance of the Cells class: Cells cells; We build this in the Memory constructor in a nested for loop, where each cell is successively instantiated with a row and column, given a random color, and added to cells: Memory(this.length) { cells = new Cells(); var cell; for (var x = 0; x < length; x++) { for (var y = 0; y < length; y++) { cell = new Cell(this, x, y); cell.color = randomColor(); cells.add(cell); } } } We can draw a rectangle and fill it with a color at the same time. So, we realize that we don't need to draw lines as we did in the previous spiral! The _boxes method is called from the draw animation: with a for…in statement, we loop over the collection of cells and call the _colorBox method that will draw and color the cell for each cell: void _boxes() { for (Cell cell in memory.cells) { _colorBox(cell); } } void _colorBox(Cell cell) { var gap = height / memory.length; var x = cell.row * gap; var y = cell.column * gap; context ..beginPath() ..fillStyle = colorMap[cell.color] ..rect(x, y, gap, gap) ..fill() ..stroke() ..closePath(); } Spiral 4 – implementing the rules However, wait! Our game can only work if the same color appears in only two cells: a cell and its twin cell. Moreover, a cell can be hidden or not: the color can be seen or not? To take care of this, the Cell class gets two new attributes: Cell twin; bool hidden = true; The _colorBox method in the Board class can now show the color of the cell when hidden is false (line (2)); when hidden = true (the default state), a neutral gray color will be used for the cell (line (1)): static const String COLOR_CODE = '#f0f0f0'; We also gave the gap variable a better name, boxSize: void _colorBox(Cell cell) { var x = cell.column * boxSize; var y = cell.row * boxSize; context.beginPath(); if (cell.hidden) { context.fillStyle = COLOR_CODE; (1) } else { context.fillStyle = colorMap[cell.color]; (2) } // same code as in Spiral 3 } The lines (1) and (2) can also be stated more succinctly with the ? ternary operator. Remember that the drawing changes because the _colorBox method is called via draw at 60 frames per second and the board can react to a mouse click. In this spiral, we will show a cell when it is clicked together with its twin cell and then they will stay visible. Attaching an event handler for this is easy. We add the following line to the Board constructor: querySelector('#canvas').onMouseDown.listen(onMouseDown); The onMouseDown event handler has to know on which cell the click occurred. The mouse event e contains the coordinates of the click in its e.offset.x and e.offset.y properties (lines (3) and (4)). We will obtain the cell's row and column by using a truncating division ~/ operator dividing the x (which gives the column) and y (which gives the row) values by boxSize: void onMouseDown(MouseEvent e) { int row = e.offset.y ~/ boxSize; (3) int column = e.offset.x ~/ boxSize; (4) Cell cell = memory.getCell(row, column); (5) cell.hidden = false; (6) cell.twin.hidden = false; (7) } Memory has a collection of cells. To get the cell with a specified row and column value, we will add a getCell method to memory and call it in line (5). When we have the cell, we will set its hidden property and that of its twin cell to false (lines (6) to (7)). The getCell method must return the cell at the given row and column. It loops through all the cells in line (8) and checks each cell, whether it is positioned at that row and column (line (9)). If yes, it will return that cell: Cell getCell(int row, int column) { for (Cell cell in cells) { (8) if (cell.intersects(row, column)) { (9) return cell; } } } For this purpose, we will add an intersects method to the Cell class. This checks whether its row and column match the given row and column for the current cell (see line (10)): bool intersects(int row, int column) { if (this.row == row && this.column == column) { (10) return true; } return false; } Now, we have already added a lot of functionality, but the drawing of the board will need some more thinking: How to give a cell (and its twin cell) a random color that is not yet used? How to attach a cell randomly to a twin cell that is not yet used? To end this, we will have to make the constructor of Memory a lot more intelligent: Memory(this.length) { if (length.isOdd) { (1) throw new Exception( 'Memory length must be an even integer: $length.'); } cells = new Cells(); var cell, twinCell; for (var x = 0; x < length; x++) { for (var y = 0; y < length; y++) { cell = getCell(y, x); (2) if (cell == null) { (3) cell = new Cell(this, y, x); cell.color = _getFreeRandomColor(); (4) cells.add(cell); twinCell = _getFreeRandomCell(); (5) cell.twin = twinCell; (6) twinCell.twin = cell; twinCell.color = cell.color; cells.add(twinCell); } } } } The number of pairs given by ((length * length) / 2) must be even. This is only true if the length parameter of Memory itself is even, so we checked it in line (1). Again, we coded a nested loop and got the cell at that row and column. However, as the cell at that position has not yet been made (line (3)), we continued to construct it and assign its color and twin. In line (4), we called _getFreeRandomColor to get a color that is not yet used: String _getFreeRandomColor() { var color; do { color = randomColor(); } while (usedColors.any((c) => c == color)); (7) usedColors.add(color); (8) return color; } The do…while loop continues as long as the color is already in a list of usedColors. On exiting from the loop, we found an unused color, which is added to usedColors in line (8) and also returned. We then had to set everything for the twin cell. We searched for a free one with the _getFreeRandomCell method in line (5). Here, the do…while loop continues until a (row, column) position is found where cell == null is, meaning that we haven't yet created a cell there (line (9)). We will promptly do this in line (10): Cell _getFreeRandomCell() { var row, column; Cell cell; do { row = randomInt(length); column = randomInt(length); cell = getCell(row, column); } while (cell != null); (9) return new Cell(this, row, column); (10) } From line (6) onwards, the properties of the twin cell are set and added to the list. This is all we need to produce the following result: Paired colored cells Spiral 5 – game logic (bringing in the time element) Our app isn't playable yet: When a cell is clicked, its color must only show for a short period of time (say one second) When a cell and its twin cell are clicked within a certain time interval, they must remain visible All of this is coded in the mouseDown event handler and we also need a lastCellClicked variable of the Cell type in the Board class. Of course, this is exactly the cell we get in the mouseDown event handler. So, we will set it in line (5) in the following code snippet: void onMouseDown(MouseEvent e) { // same code as in Spiral 4 - if (cell.twin == lastCellClicked && lastCellClicked.shown) { (1) lastCellClicked.hidden = false; (2) if (memory.recalled) memory.hide(); (3) } else { new Timer(const Duration(milliseconds: 1000), () => cell.hidden = true); (4) } lastCellClicked = cell; (5) } In line (1), we checked whether the last clicked cell was the twin cell and whether this is still shown. Then, we made sure in (2) that it stays visible. shown is a new getter in the Cell class to make the code more readable: bool get shown => !hidden;. If at that moment all the cells were shown (the memory is recalled), we again hid them in line (3). If the last clicked cell was not the twin cell, we hid the current cell after one second in line (4). recalled is a simple getter (read-only property) in the Memory class and it makes use of a Boolean variable in Memory that is initialized to false (_recalled = false;): bool get recalled { if (!_recalled) { if (cells.every((c) => c.shown)) { (6) _recalled = true; } } return _recalled; } In line (6), we tested that if every cell is shown, then this variable is set to true (the game is over). every is a new method in the Cells List and a nice functional way to write this is given as follows: bool every(Function f) => list.every(f); The hide method is straightforward: hide every cell and reset the _recalled variable to false: hide() { for (final cell in cells) cell.hidden = true; _recalled = false; } This is it, our game works! Spiral 6 – some finishing touches A working program always gives its developer a sense of joy, and rightfully so. However, this doesn't that mean you can leave the code as it is. On the contrary, carefully review your code for some time to see whether there is room for improvement or optimization. For example, are the names you used clear enough? The color of a hidden cell is now named simply COLOR_CODE in board.dart, renaming it to HIDDEN_CELL_COLOR_CODE makes its meaning explicit. The List object used in the Cells class can indicate that it is List<Cell>, by applying the fact that Dart lists are generic. The parameter of the every method in the Cell class is more precise—it is a function that accepts a cell and returns bool. Our onMouseDown event handler contains our game logic, so it is very important to tune it if possible. After some thought, we see that the code from the previous spiral can be improved; in the following line, the second condition after && is, in fact, unnecessary: if (cell.twin == lastCellClicked && lastCellClicked.shown) {...} When the player has guessed everything correctly, showing the completed screen for a few seconds will be more satisfactory (line (2)). So, this portion of our event handler code will change to: if (cell.twin == lastCellClicked) { (1) lastCellClicked.hidden = false; if (memory.recalled) { // game over new Timer(const Duration(milliseconds: 5000), () => memory.hide()); (2) } } else if (cell.twin.hidden) { new Timer(const Duration(milliseconds: 800), () => cell.hidden = true); } Why don’t we show a "YOU HAVE WON!" banner. We will do this by drawing the text on the canvas (line (3)), so we must do it in the draw() method (otherwise, it would disappear after INTERVAL milliseconds): void draw() { _clear(); _boxes(); if (memory.recalled) { // game over context.font = "bold 25px sans-serif"; context.fillStyle = "red"; context.fillText("YOU HAVE WON !", boxSize, boxSize * 2); (3) } } Then, the same game with the same configuration can be played again. We could make it more obvious that a cell is hidden by decorating it with a small circle in the _colorBox method (line (4)): if (cell.hidden) { context.fillStyle = HIDDEN_CELL_COLOR_CODE; var centerX = cell.column * boxSize + boxSize / 2; var centerY = cell.row * boxSize + boxSize / 2; var radius = 4; context.arc(centerX, centerY, radius, 0, 2 * PI, false); (4) } We do want to give our player a chance to start over by supplying a Play again button. The easiest way will be to simply refresh the screen (line (5)) by adding this code to the startup script: void main() { canvas = querySelector('#canvas'); ButtonElement play = querySelector('#play'); play.onClick.listen(playAgain); new Board(canvas, new Memory(4)); } playAgain(Event e) { window.location.reload(); (5) } Spiral 7 – using images One improvement that certainly comes to mind is the use of pictures instead of colors as shown in the Using images screenshot. How difficult would that be? It turns out that this is surprisingly easy, because we already have the game logic firmly in place! In the images folder, we supply a number of game pictures. Instead of the color property, we give the cell a String property (image), which will contain the name of the picture file. We then replace utilcolor.dart with utilimages.dart, which contains a imageList variable with the image filenames. In utilrandom.dart, we will replace the color methods with the following code: String randomImage() => randomListElement(imageList); The changes to memory.dart are also straightforward: replace the usedColor list with List usedImages = []; and the _getFreeRandomColor method with _getFreeRandomImage, which will use the new list and method: List usedImages = []; String _getFreeRandomImage() { var image; do { image = randomImage(); } while (usedImages.any((i) => i == image)); usedImages.add(image); return image; } In board.dart, we replace _colorBox(cell) with _imageBox(cell). The only new thing is how to draw the image on canvas. For this, we need ImageElement objects. Here, we have to be careful to create these objects only once and not over and over again in every draw cycle, because this produces a flickering screen. We will store the ImageElements object in a Map: var imageMap = new Map<String, ImageElement>(); Then, we populate this in the Board constructor with a for…in loop over memory.cells: for (var cell in memory.cells) { ImageElement image = new Element.tag('img'); (1) image.src = 'images/${cell.image}'; (2) imageMap[cell.image] = image; (3) } We create a new ImageElement object in line (1), giving it the complete file path to the image file as a src property in line (2) and store it in imageMap in line (3). The image file will then be loaded into memory only once. We don't do any unnecessary network access to effectively cache the images. In the draw cycle, we will load the image from imageMap and draw it in the current cell with the drawImage method in line (4): if (cell.hidden) { // see previous code } else { ImageElement image = imageMap[cell.image]; context.drawImage(image, x, y); // resize to cell size (4) } Perhaps, you can think of other improvements? Why not let the player specify the game difficulty by asking the number of boxes. It is 16 now. Check whether the input is a square of an even number. Do you have enough colors to choose from? Perhaps, dynamically building a list with enough random colors would be a better idea. Calculating and storing the statistics discussed in the model would also make the game more attractive. Another enhancement from the model is to support different catalogs of pictures. Go ahead and exercise your Dart skills! Summary By thoroughly investigating two games applying all of Dart we have already covered, your Dart star begins to shine. For other Dart games, visit http://www.builtwithdart.com/projects/games/. You can find more information at http://www.dartgamedevs.org/ on building games. Resources for Article: Further resources on this subject: Slideshow Presentations [article] Dart with JavaScript [article] Practical Dart [article]
Read more
  • 0
  • 0
  • 19049

article-image-gns3-orchestra
Packt
29 Oct 2013
9 min read
Save for later

The GNS3 orchestra

Packt
29 Oct 2013
9 min read
(For more resources related to this topic, see here.) You will observe multiple TCP connections and UDP pipes being created from both the GNS3 management console and your operating system's command line. To get a closer look at how the conductor works, open GNS3 to a new blank canvas and issue the command debug 3 in the GNS3 command console : => debug 3 As you open GNS3, the conductor readies the players awaiting your instructions. The moment you drag your first Cisco router onto the workspace, GNS3 spawns an instance of Dynamips and connects to it on port 7200. You can see this in two places: To reproduce the effects shown here, use a C7200 router image with 256MB RAM. By issuing a netstat–a command in a Windows/Linux/OS X command window: C: >netstat -an ... Proto Local Address Foreign Address State TCP 127.0.0.1:7200 0.0.0.0:0 LISTENING TCP 127.0.0.1:7200 127.0.0.1:49194 ESTABLISHED TCP 127.0.0.1:49194 127.0.0.1:7200 ESTABLISHED In the output of the GNS3 management console. The following output lines are abbreviated to conserve space: Hypervisor manager: connecting on 127.0.0.1:7200 Hypervisor manager: connected to hypervisor on 127.0.0.1 port 7200 Also in the output of the GNS3 management console , note the following (trimmed) lines: Hypervisor manager: hypervisor base UDP is 10001 …<snip>… PORT TRACKER: allocate port 2101 sending to dynamips at 127.0.0.1:7200 -> vm set_con_tcp_port R1 2101 returned -> ['100-OK'] PORT TRACKER: allocate port 2501 sending to dynamips at 127.0.0.1:7200 -> vm set_aux_tcp_port R1 2501 returned -> ['100-OK'] The UDPbase port I'll deal with shortly, but notice that GNS3 has told Dynamips to prepare to open ports 2101 and 2501 for console and AUX port communications respectively, which are the base ports defined in GNS3 Preferences for the Dynamips setting under the Dynamips tab. Also note that these ports are not yet opened, in the orchestral analogy you could say they are merely being tuned up at this stage. Next, add a second router and observe (in the console output) that the console and AUX port allocations have incremented by one, but there is no change to the base UDP port. sending to dynamips at 127.0.0.1:7200 -> vm set_con_tcp_port R2 2102 sending to dynamips at 127.0.0.1:7200 -> vm set_aux_tcp_port R2 2502 You are about to connect these two devices. Configure them with FastEthernet interfaces if necessary, or use the FastEthernet Add a link tool, and connect the two routers. Watch the console output for these lines: Connect link from R1 f1/0 to R2 f1/0 new base UDP port for dynamips at 127.0.0.1:7200 is now: 10002 new base UDP port for dynamips at 127.0.0.1:7200 is now: 10003 sending to dynamips at 127.0.0.1:7200 -> nio create_udp nio_udp0 10001 127.0.0.1 10002 sending to dynamips at 127.0.0.1:7200 -> nio create_udp nio_udp1 10002 127.0.0.1 10001 sending to dynamips at 127.0.0.1:7200 -> vm slot_add_nio_binding R1 1 0 nio_udp0 sending to dynamips at 127.0.0.1:7200 -> vm slot_add_nio_binding R2 1 0 nio_udp1 And on your host computer, netstat -an reveals: C:> netstat -an | find "1000" UDP 0.0.0.0:10001 *:* UDP 0.0.0.0:10002 *:* Understanding what is going on here is the key to understanding how Dynamips achieves communication between routers. What has just happened is that a UDP tunnel has been created between these two devices. UDP tunnel concept Links between devices in GNS3 is achieved using UDP tunnels. What this means in this scenario is that whenever R1 sends a frame from interface f1/0, the entire frame, including the Source MAC address, destination MAC address and payload, gets put inside a UDP packet with a source port of 10001 and a destination IP address:destination port of 127.0.0.1:10002 which means that the frame will end up at R2's f1/0 interface because it is bound to port 10002. The return frames take the reverse path: source port 10002, destination IP:port 127.0.0.1:10001. To illustrate this, I assigned an IP addresses of 1.1.1.1 and 1.1.1.2 to interface f1/0 on R1 and R2 respectively, then captured a ping packet on the link between R1 and R2 on the host computer's loopback interface. The Wireshark capture shown in the following screenshot shows a ping packet from 1.1.1.1 on its way to 1.1.1.2, but you can see that the entire layer 2 frame (1), including the layer 2 MAC addresses of R1 and R2 is encapsulated inside a UDP packet travelling from 127.0.0.1:10001 to 127.0.0.1:10002 (2). Another thing to note is that the first UDP port used was the Base UDP port defined in GNS3 Preferences, Dynamips settings under the Dynamips tab. Now would also be a good time to issue a show run command in the GNS3 management console window, to see how GNS3 is building up your topology.net file. => show runautostart = False [127.0.0.1:7200] workingdir = C:UserschrisAppDataLocalTempGNS3_rwftbworking udp = 10001 [[7200]] image = C:UserschrisGNS3Imagesc7200-p-mz.124-10a.image ram = 256 idlepc = 0x60750000 sparsemem = True ghostios = True [[ROUTER R1]] console = 2101 aux = 2501 slot1 = PA-2FE-TX f1/0 = R2 f1/0 [[ROUTER R2]] console = 2102 aux = 2502 slot1 = PA-2FE-TX f1/0 = R1 f1/0 Note that the amount of RAM set for each of these routers is 256MiB. Also recall that in the Hypervisor Manager settings previously shown, the Memory limit per hypervisor was set to 512MiB. Now add another router, and watch the console output, and check your host computer's TCP connections again with the netstat -an command. You will see of course: Hypervisor manager: connecting on 127.0.0.1:7201 and… TCP 127.0.0.1:7201 0.0.0.0:0 LISTENING This shows that a second hypervisor instance has been created, and allocated TCP port 7201 for communication. You will also see this reflected in the configuration information if you issue another a show run command in the GNS3 management console window. ...<Output omitted>... [127.0.0.1:7201] workingdir = C:UserschrisAppDataLocalTempGNS3_rwftbworking udp = 10101 This also reveals that the base UDP port for this hypervisor is 10101, recall that the value for the UDP incrementation in the Dynamips Hypervisor Manager setting page was 100, so the base UDP port for this instance of the hypervisor is 100 greater than the general base UDP port of 10001 for Dynamips. You can probably predict what UDP port numbers will be used then if you now connect R2 to R3 with a FastEthernet link. Make the link and see if your prediction was correct: sending to dynamips at 127.0.0.1:7200 -> nio create_udp nio_udp2 10003 127.0.0.1 10101 sending to dynamips at 127.0.0.1:7201 -> nio create_udp nio_udp3 10101 127.0.0.1 10003 sending to dynamips at 127.0.0.1:7200 -> vm slot_add_nio_binding R2 1 1 nio_udp2 sending to dynamips at 127.0.0.1:7201 -> vm slot_add_nio_binding R3 1 0 nio_udp3 Did you predict that the next connection would be made from port 10003 to 10101? Well done. But what if you add a switch or a hub? Add a generic Ethernet switch to the topology, and issue a show run command in the GNS3 management console window. You will notice that there is NO reference to the switch in the output, and in fact if you saved your topology at this point and loaded it later, there would be no switch in your topology. That is because the switch doesn't get allocated to a hypervisor until it has at least one connection to another item in the topology. The question is, since our topology has two hypervisors running, which hypervisor will be allocated the switch? Connect your recently added switch to R1. Observe what happens in the GNS3 management console, and issue another show run command in the GNS3 management console window. Here is what you are looking for: Firstly, you should see the connections being created. Note that the UDP port numbers are from the range allocated to the first hypervisor that was spawned, NOT the most recent hypervisor spawned. sending to dynamips at 127.0.0.1:7200 -> nio create_udp nio_udp4 10004 127.0.0.1 10005 sending to dynamips at 127.0.0.1:7200 -> nio create_udp nio_udp5 10005 127.0.0.1 10004 Secondly, in the topology description, you can see that SW1 has been assigned to the hypervisor running on TCP port 7200 which allocated UDP ports from the 10000+ range. What actually happens is that GNS3 assigns generic devices like switches, hubs, and clouds to the hypervisor to which the device is first connected. => show run autostart = False [127.0.0.1:7200] workingdir = C:UserschrisAppDataLocalTemp udp = 10001 [[7200]] ... [[ETHSW SW1]] 1 = access 1 R1 f1/1 [[ROUTER R1]] ... And finally, if you changed your Memory usage per hypervisor setting back on page 73, don't forget to change it back. I recommend setting it to 1024MiB . By now you are probably wondering how GNS3 and Dynamips deal with the other supported emulators: Qemu and Oracle VirtualBox . Conducting Qemu and VirtualBox Recall that Dynamips is a hypervisor used to initiate the spawning of Cisco router VMs (virtual machines) instances, and configure host communication to these VMs via console AUX and virtual network interfaces. Just like Dynamips, both Qemu and Oracle VirtualBox follow a similar hypervisor model, only in this case the hypervisor is "wrapped" to give it similar functionality to Dynamips. The wrappers for the hypervisors are called qemuwrapper and vboxwrapper respectively, and these wrappers listen on ports 10525 and 11525 as shown in the configuration options in GNS3 Preferences under the Qemu and VirtualBox settings. Trivia: Port 10525 was chosen by Thomas Pani when he wrote pemuwrapper, the wrapper for the PIX 525 emulator, port 1525 was already assigned by the IANA. Pemuwrapper evolved into qemuwrapper, and when Alexey Eromenko, alias "Technologov" wrote vboxwrapper he simply added 1000 to the qemuwrapper port number. Again, just like Dynamips, you can run qemuwrapper and vboxwrapper as standalone applications, then telnet to port 127.0.0.1:10525 or 127.0.0.1:11525 to issue commands like qemu version or vbox version or the commands to spawn a virtual machine if you bothered to learn the syntax. Again, like Dynamips, qemuwrapper and vboxwrapper direct the TCP port to be used for console connections and UDP port for UDP tunnel connections, the base values for these can also be found in GNS3 Preferences under the Qemu and VirtualBox settings. But unlike Dynamips, the wrapper is NOT the hypervisor as well. The hypervisor is Qemu or VirtualBox, so these applications had to be compiled to allow communication via UDP tunnel interfaces. In the case of Qemu prior to Version 1.1, this required a specially compiled version. The GNS3 downloads page has links to the patched Version 0.11 that I used throughout this article. VirtualBox has built-in support for UDP tunnels. To see the full GNS3 orchestra playing, you can now add Qemu and VirtualBox devices to your topology and watch the GNS3 management console and check your TCP/UDP connections with the netstat -an command. As you watch the GNS3 management console you will see the hidden power of GNS3 beyond the GNS3 GUI as it conducts its orchestral sections of Dynamips, qemuwrapper, and vboxwrapper to play in harmony, and even see that they have their own sections in the GNS3 topology.net file, as can be seen by issuing a show run command in the GNS3 management console . Summary This article covers all we need to know about the GNS3 Orcehstra—the GNS3 GUI Resources for Article: Further resources on this subject: Editing attributes [Article] iPhone JavaScript: Installing Frameworks [Article] Visualizing my Social Graph with d3.js [Article]
Read more
  • 0
  • 0
  • 19040
article-image-introduction-creational-patterns-using-go-programming
Packt
02 Jan 2017
12 min read
Save for later

Introduction to Creational Patterns using Go Programming

Packt
02 Jan 2017
12 min read
This article by Mario Castro Contreras, author of the book Go Design Patterns, introduces you to the Creational design patterns that are explained in the book. As the title implies, this article groups common practices for creating objects. Creational patterns try to give ready-to-use objects to users instead of asking for their input, which, in some cases, could be complex and will couple your code with the concrete implementations of the functionality that should be defined in an interface. (For more resources related to this topic, see here.) Singleton design pattern – Having a unique instance of an object in the entire program Have you ever done interviews for software engineers? It's interesting that when you ask them about design patterns, more than 80% will start saying Singleton design pattern. Why is that? Maybe it's because it is one of the most used design patterns out there or one of the easiest to grasp. We will start our journey on creational design patterns because of the latter reason. Description Singleton pattern is easy to remember. As the name implies, it will provide you a single instance of an object, and guarantee that there are no duplicates. At the first call to use the instance, it is created and then reused between all the parts in the application that need to use that particular behavior. Objective of the Singleton pattern You'll use Singleton pattern in many different situations. For example: When you want to use the same connection to a database to make every query When you open a Secure Shell (SSH) connection to a server to do a few tasks, and don't want to reopen the connection for each task If you need to limit the access to some variable or space, you use a Singleton as the door to this variable. If you need to limit the number of calls to some places, you create a Singleton instance to make the calls in the accepted window The possibilities are endless, and we have just mentioned some of them. Implementation Finally, we have to implement the Singleton pattern. You'll usually write a static method and instance to retrieve the Singleton instance. In Go, we don't have the keyword static, but we can achieve the same result by using the scope of the package. First, we create a structure that contains the object which we want to guarantee to be a Singleton during the execution of the program: package creational type singleton struct{ count int } var instance *singleton func GetInstance() *singleton { if instance == nil { instance = new(singleton) } return instance } func (s *singleton) AddOne() int { s.count++ return s.count } We must pay close attention to this piece of code. In languages like Java or C++, the variable instance would be initialized to NULL at the beginning of the program. In Go, you can initialize a pointer to a structure as nil, but you cannot initialize a structure to nil (the equivalent of NULL). So the var instance *singleton line defines a pointer to a structure of type Singleton as nil, and the variable called instance. We created a GetInstance method that checks if the instance has not been initialized already (instance == nil), and creates an instance in the space already allocated in the line instance = new(singleton). Remember, when we use the keyword new, we are creating a pointer to the type between the parentheses. The AddOne method will take the count of the variable instance, raise it by one, and return the current value of the counter. Lets run now our unit tests again: $ go test -v -run=GetInstance === RUN TestGetInstance --- PASS: TestGetInstance (0.00s) PASS ok Factory method – Delegating the creation of different types of payments The Factory method pattern (or simply, Factory) is probably the second-best known and used design pattern in the industry. Its purpose is to abstract the user from the knowledge of the structure it needs to achieve a specific purpose. By delegating this decision to a Factory, this Factory can provide the object that best fits the user needs or the most updated version. It can also ease the process of downgrading or upgrading of the implementation of an object if needed. Description When using the Factory method design pattern, we gain an extra layer of encapsulation so that our program can grow in a controlled environment. With the Factory method, we delegate the creation of families of objects to a different package or object to abstract us from the knowledge of the pool of possible objects we could use. Imagine that you have two ways to access some specific resource: by HTTP or FTP. For us, the specific implementation of this access should be invisible. Maybe, we just know that the resource is in HTTP or in FTP, and we just want a connection that uses one of these protocols. Instead of implementing the connection by ourselves, we can use the Factory method to ask for the specific connection. With this approach, we can grow easily in the future if we need to add an HTTPS object. Objective of the Factory method After the previous description, the following objectives of the Factory Method design pattern must be clear to you: Delegating the creation of new instances of structures to a different part of the program Working at the interface level instead of with concrete implementations Grouping families of objects to obtain a family object creator Implementation We will start with the GetPaymentMethod method. It must receive an integer that matches with one of the defined constants of the same file to know which implementation it should return. package creational import ( "errors" "fmt" ) type PaymentMethod interface { Pay(amount float32) string } const ( Cash = 1 DebitCard = 2 ) func GetPaymentMethod(m int) (PaymentMethod, error) { switch m { case Cash: return new(CashPM), nilcase DebitCard: return new(DebitCardPM), nil default: return nil, errors.New(fmt.Sprintf("Payment method %d not recognizedn", m)) } } We use a plain switch to check the contents of the argument m (method). If it matches any of the known methods—cash or debit card, it returns a new instance of them. Otherwise, it will return a nil and an error indicating that the payment method has not been recognized. Now we can run our tests again to check the second part of the unit tests: $go test -v -run=GetPaymentMethod . === RUN TestGetPaymentMethodCash --- FAIL: TestGetPaymentMethodCash (0.00s) factory_test.go:16: The cash payment method message wasn't correct factory_test.go:18: LOG: === RUN TestGetPaymentMethodDebitCard --- FAIL: TestGetPaymentMethodDebitCard (0.00s) factory_test.go:28: The debit card payment method message wasn't correct factory_test.go:30: LOG: === RUN TestGetPaymentMethodNonExistent --- PASS: TestGetPaymentMethodNonExistent (0.00s) factory_test.go:38: LOG: Payment method 20 not recognized FAIL exit status 1 FAIL Now we do not get the errors saying it couldn't find the type of payment methods. Instead, we receive a message not correct error when it tries to use any of the methods that it covers. We also got rid of the Not implemented message that was being returned when we asked for an unknown payment method. Lets implement the structures now: type CashPM struct{} type DebitCardPM struct{} func (c *CashPM) Pay(amount float32) string { return fmt.Sprintf("%0.2f paid using cashn", amount) } func (c *DebitCardPM) Pay(amount float32) string { return fmt.Sprintf("%#0.2f paid using debit cardn", amount) } We just get the amount, printing it in a nice formatted message. With this implementation, the tests will all passing now: $ go test -v -run=GetPaymentMethod . === RUN TestGetPaymentMethodCash --- PASS: TestGetPaymentMethodCash (0.00s) factory_test.go:18: LOG: 10.30 paid using cash === RUN TestGetPaymentMethodDebitCard --- PASS: TestGetPaymentMethodDebitCard (0.00s) factory_test.go:30: LOG: 22.30 paid using debit card === RUN TestGetPaymentMethodNonExistent --- PASS: TestGetPaymentMethodNonExistent (0.00s) factory_test.go:38: LOG: Payment method 20 not recognized PASS ok Do you see the LOG: messages? They aren't errors—we just print some information that we receive when using the package under test. These messages can be omitted unless you pass the -v flag to the test command: $ go test -run=GetPaymentMethod . ok Abstract Factory – A factory of factories After learning about the factory design pattern is when we grouped a family of related objects in our case payment methods, one can be quick to think: what if I group families of objects in a more structured hierarchy of families? Description The Abstract Factory design pattern is a new layer of grouping to achieve a bigger (and more complex) composite object, which is used through its interfaces. The idea behind grouping objects in families and grouping families is to have big factories that can be interchangeable and can grow more easily. In the early stages of development, it is also easier to work with factories and abstract factories than to wait until all concrete implementations are done to start your code. Also, you won't write an Abstract Factory from the beginning unless you know that your object's inventory for a particular field is going to be very large and it could be easily grouped into families. The objective Grouping related families of objects is very convenient when your object number is growing so much that creating a unique point to get them all seems the only way to gain flexibility of the runtime object creation. Following objectives of the Abstract Factory method must be clear to you: Provide a new layer of encapsulation for Factory methods that returns a common interface for all factories Group common factories into a super Factory (also called factory of factories) Implementation The implementation of every factory is already done for the sake of brevity. They are very similar to the factory method with the only difference being that in the factory method, we don't use an instance of the factory, because we use the package functions directly. The implementation of the vehicle factory is as follows: func GetVehicleFactory(f int) (VehicleFactory, error) { switch f { case CarFactoryType: return new(CarFactory), nil case MotorbikeFactoryType: return new(MotorbikeFactory), nil default: return nil, errors.New(fmt.Sprintf("Factory with id %d not recognizedn", f)) } } Like in any factory, we switched between the factory possibilities to return the one that was demanded. As we have already implemented all concrete vehicles, the tests must be run too: go test -v -run=Factory -cover . === RUN TestMotorbikeFactory --- PASS: TestMotorbikeFactory (0.00s) vehicle_factory_test.go:16: Motorbike vehicle has 2 wheels vehicle_factory_test.go:22: Sport motorbike has type 1 === RUN TestCarFactory --- PASS: TestCarFactory (0.00s) vehicle_factory_test.go:36: Car vehicle has 4 seats vehicle_factory_test.go:42: Luxury car has 4 doors. PASS coverage: 45.8% of statements ok All of them passed. Take a close look and note that we have used the -cover flag when running the tests to return a coverage percentage of the package 45.8%. What this tells us is that 45.8% of the lines are covered by the tests we have written, but 54.2% is still not under the tests. This is because we haven't covered the cruise motorbike and the Family car with tests. If you write those tests, the result should rise to around 70.8%. Prototype design pattern The last pattern we will see in this article is the Prototype pattern. Like all creational patterns, this too comes in handy when creating objects and it is very common to see the Prototype pattern surrounded by more patterns. Description The aim of the Prototype pattern is to have an object or a set of objects that are already created at compilation time, but which you can clone as many times as you want at runtime. This is useful, for example, as a default template for a user who has just registered with your webpage or a default pricing plan in some service. The key difference between this and a Builder pattern is that objects are cloned for the user instead of building them at runtime. You can also build a cache-like solution, storing information using a prototype. Objective Maintain a set of objects that will be cloned to create new instances Free CPU of complex object initialization to take more memory resources We will start with the GetClone method. This method should return an item of the specified type: type ShirtsCache struct {} func (s *ShirtsCache)GetClone(m int) (ItemInfoGetter, error) { switch m { case White: newItem := *whitePrototype return &newItem, nil case Black: newItem := *blackPrototype return &newItem, nil case Blue: newItem := *bluePrototype return &newItem, nil default: return nil, errors.New("Shirt model not recognized") } } The Shirt structure also needs a GetInfo implementation to print the contents of the instances. type ShirtColor byte type Shirt struct { Price float32 SKU string Color ShirtColor } func (s *Shirt) GetInfo() string { return fmt.Sprintf("Shirt with SKU '%s' and Color id %d that costs %fn", s.SKU, s.Color, s.Price) } Finally, lets run the tests to see that everything is now working: go test -run=TestClone -v . === RUN TestClone --- PASS: TestClone (0.00s) prototype_test.go:41: LOG: Shirt with SKU 'abbcc' and Color id 1 that costs 15.000000 prototype_test.go:42: LOG: Shirt with SKU 'empty' and Color id 1 that costs 15.000000 prototype_test.go:44: LOG: The memory positions of the shirts are different 0xc42002c038 != 0xc42002c040 PASS ok In the log (remember to set the -v flag when running the tests), you can check that shirt1 and shirt2 have different SKUs. Also, we can see the memory positions of both objects. Take into account that the positions shown on your computer will probably be different. Summary We have seen the creational design patterns commonly used in the software industry. Their purpose is to abstract the user from the creation of objects for handling complexity or maintainability purposes. Design patterns have been the foundation of thousands of applications and libraries since the nineties, and most of the software we use today has many of these creational patterns under the hood. Resources for Article: Further resources on this subject: Getting Started [article] Thinking Functionally [article] Auditing and E-discovery [article]
Read more
  • 0
  • 0
  • 19031

article-image-ai-now-institute-publishes-a-report-on-the-diversity-crisis-in-ai-and-offers-12-solutions-to-fix-it
Bhagyashree R
22 Apr 2019
7 min read
Save for later

AI Now Institute publishes a report on the diversity crisis in AI and offers 12 solutions to fix it

Bhagyashree R
22 Apr 2019
7 min read
Earlier this month, the AI Now Institute published a report, authored by Sarah Myers West, Meredith Whittaker, and Kate Crawford, highlighting the link between the diversity issue in the current AI industry and the discriminating behavior of AI systems. The report further recommends some solutions to these problems that companies and the researchers behind these systems need to adopt to address these issues. Sarah Myers West is a postdoc researcher at the AI Now Institute and an affiliate researcher at the Berkman-Klein Center for Internet and Society. Meredith Whittaker is the co-founder of the AI Now Institute and leads Google's Open Research Group and the Google Measurement Lab.  Kate Crawford is a Principal Researcher at Microsoft Research and the co-founder and Director of Research at the AI Now Institute. Kate Crawford tweeted about this study. https://twitter.com/katecrawford/status/1118509988392112128 The AI industry lacks diversity, gender neutrality, and bias-free systems In recent years, we have come across several cases of “discriminating systems”. Facial recognition systems miscategorize black people and sometimes fails to work for trans drivers. When trained in online discourse, chatbots easily learn racist and misogynistic language. This type of behavior by machines is actually a reflection of society. “In most cases, such bias mirrors and replicates existing structures of inequality in the society,” says the report. The study also sheds light on gender bias in the current workforce. According to the report, only 18% of authors at some of the biggest AI conferences are women. On the other side of the spectrum are men who cover 80%. The tech giants, Facebook and Google, have a meager 15% and 10% women as their AI research staff. The situation for black workers in the AI industry looks even worse. While Facebook and Microsoft have 4% of their current workforce as black workers, Google stands at just 2.5%. Also, vast majority of AI studies assume gender is binary, and commonly assigns people as ‘male’ or ‘female’ based on physical appearance and stereotypical assumptions, erasing all other forms of gender identity. The report further reveals that, though there have been various “pipeline studies” to check the flow of diverse job candidates, they have failed to show substantial progress in bringing diversity in the AI industry. “The focus on the pipeline has not addressed deeper issues with workplace cultures, power asymmetries, harassment, exclusionary hiring practices, unfair compensation, and tokenization that are causing people to leave or avoid working in the AI sector altogether,” the report reads. What steps can industries take to address bias and discrimination in AI Systems The report lists 12 recommendations that AI researchers and companies should employ to improve workplace diversity and address bias and discrimination in AI systems. Publish compensation levels, including bonuses and equity, across all roles and job categories, broken down by race and gender. End pay and opportunity inequality, and set pay and benefit equity goals that include contract workers, temps, and vendors. Publish harassment and discrimination transparency reports, including the number of claims over time, the types of claims submitted, and actions taken. Change hiring practices to maximize diversity: include targeted recruitment beyond elite universities, ensure more equitable focus on under-represented groups, and create more pathways for contractors, temps, and vendors to become full-time employees. Commit to transparency around hiring practices, especially regarding how candidates are leveled, compensated, and promoted. Increase the number of people of color, women and other under-represented groups at senior leadership levels of AI companies across all departments. Ensure executive incentive structures are tied to increases in hiring and retention of underrepresented groups. For academic workplaces, ensure greater diversity in all spaces where AI research is conducted, including AI-related departments and conference committees. Remedying bias in AI systems is almost impossible when these systems are opaque. Transparency is essential, and begins with tracking and publicizing where AI systems are used, and for what purpose. Rigorous testing should be required across the lifecycle of AI systems in sensitive domains. Pre-release trials, independent auditing, and ongoing monitoring are necessary to test for bias, discrimination, and other harms. The field of research on bias and fairness needs to go beyond technical debiasing to include a wider social analysis of how AI is used in context. This necessitates including a wider range of disciplinary expertise. The methods for addressing bias and discrimination in AI need to expand to include assessments of whether certain systems should be designed at all, based on a thorough risk assessment. AI-related departments and conference committees. Credits: AI Now Institute Bringing diversity in the AI workforce In order to address the diversity issue in the AI industry, companies need to make changes in the current hiring practices. They should have a more equitable focus on under-represented groups. People of color, women, and other under-represented groups should get fair chance to get into senior leadership level of AI companies across all departments. Further opportunities should be created for contractors, temps, and vendors to become full-time employees. To bridge the gender pay gap in the AI industry, it is important that companies maintain transparency regarding the compensation levels, including bonuses and equity, regardless of gender or race. In the past few years, several cases of sexual misconducts involving some of the biggest companies like Google, Microsoft, have come into light because of movements like #MeToo, Google Walkout, and more. These movements gave the victims and other supporting employees  the courage to speak against employees at higher positions who were taking undue advantage of their power. There are cases were the sexual harassment complaints were not taken seriously by the HRs and victims were told to just “get over it”. This is why, companies should  publish harassment and discrimination transparency reports that include information like the number and types of claims made and the actions taken by the company. Academic workplaces should ensure diversity in all AI-related departments and conference committees. In the past, some of the biggest AI conferences like Neural Information Processing Systems conference has failed to provide a welcoming and safer environment for women. In a survey conducted last year, many respondents shared that they have experienced sexual harassment. Women reported persistent advances from men at the conference. The organizers of such conferences should ensure an inclusive and welcoming environment for everyone. Addressing bias and discrimination in AI systems To address bias and discrimination in AI systems, the report recommends to do rigorous testing across the lifecycle of these systems. These systems should have pre-release trials, independent auditing, and monitoring to check bias, discrimination, and other harms. Looking at the social implications of AI systems, just addressing the algorithmic bias is not enough. “The field of research on bias and fairness needs to go beyond technical debiasing to include a wider social analysis of how AI is used in context. This necessitates including a wider range of disciplinary expertise,” says the report. While assessing a AI system, researchers and developers should also check whether designing a certain system is required at all, considering the risks it poses. The study calls for re-evaluating the current AI systems used for classifying, detecting, and predicting the race and gender. The idea of identifying a race or gender just by appearance is flawed and can be easily abused. Especially, systems that use physical appearance to find interior states, for instance, those that claim to detect sexuality from headshots. These systems are urgently in need to be checked. To know more in detail, read the full report: Discriminating Systems. Microsoft’s #MeToo reckoning: female employees speak out against workplace harassment and discrimination Desmond U. Patton, Director of SAFElab shares why AI systems should be a product of interdisciplinary research and diverse teams Google’s Chief Diversity Officer, Danielle Brown resigns to join HR tech firm Gusto
Read more
  • 0
  • 0
  • 19011

article-image-architecture-freeswitch
Packt
14 Jun 2013
13 min read
Save for later

Architecture of FreeSWITCH

Packt
14 Jun 2013
13 min read
(For more resources related to this topic, see here.) A revolution has begun and secrets have been revealed How and why the telephone works is a mystery to most people. It has been kept secret for years. We just plugged our phones into the wall and they worked, and most people do just that and expect it to work. The telephony revolution has begun, and we have begun to pry its secrets from the clutches of the legacy of the telephony industry. Now, everyday individuals like you and me are able to build phone systems that outperform traditional phone services and offer advanced features for relatively low cost. Some people even use FreeSWITCH to provide telephone services for making a profit. FreeSWITCH has been designed to make all of this easier, so we will go over the architecture to get a better understanding of how it works. Do not be concerned if some of the concepts we introduce seem unnaturally abstract. Learning telephony takes time, especially VoIP. In fact, we recommend that you read this article more than once. Absorb as much as you can on the first pass. You will be surprised at how much your understanding of VoIP and FreeSWITCH has improved.Give yourself plenty of time to digest all of these strange new concepts, and soon you will find that you are a skilled FreeSWITCH administrator. If you keep at it, you will be rewarded with a meaningful understanding of this strange and wonderful world we call telephony. Telephones and telephony systems (such as telephone switches and PBXs) are very complicated and have evolved over the years into several varieties. The most popular type of phone in the U.K. and the U.S. is the traditional analog phone, which we affectionately refer to as POTS lines or Plain Old Telephone Service. From the traditional Ma Bell phone up to the long-range cordless phones that most of us have today, one thing has remained the same—the underlying technology. In the last 10-15 years, there has been a convergence of technology between computers and telephones that has produced a pair of affordable alternatives to POTS lines—Mobile phones and VoIP phones (also called Internet Phones). FreeSWITCH fits into this big tangled mess of various telephone technologies by bridging them together, so that they can communicate despite being otherwise completely incompatible. FreeSWITCH also bridges telephone calls with computer programs that you can write yourself, and controls what happens in ways like never before. FreeSWITCH is software that runs on Windows and several UNIX varieties such as Mac OS X, Linux, Solaris, and BSD. This means you can install FreeSWITCH on your home PC or even a high-end server and use it to process phone calls. The FreeSWITCH design – modular, scalable, and stable The design goal of FreeSWITCH is to provide a modular, scalable system around a stable switching core, and to provide a robust interface for developers to add to and control the system. Various elements in FreeSWITCH are independent of each other and do not have much knowledge about how the other parts are working, other than what is provided in what are called exposed functions. The functionality of FreeSWITCH can also be extended with loadable modules, which tie a particular external technology into the core. FreeSWITCH has many different module types that revolve around the central core, much like satellites orbiting a planet. The list includes: Module type: Purpose: Endpoint Telephone protocols like SIP/H.323 and POTS lines Application Performs a task such as playing audio or setting data Application Programming Interface (API) Exports a function that takes text input and returns text output, which could be used across modules or from an external connection Automated Speech Recognition (ASR) Interfaces with speech recognition systems Chat Bridges and exchanges various chat protocols Codec Translates between audio formats Dialplan Parses the call details and decides where to route the call Directory Connects directory information services, such as LDAP, to a common core lookup API Event handlers Allows external programs to control FreeSWITCH File Provides an interface to extract and play sound from various audio file formats Formats Plays audio files in various formats Languages Programming language interfaces used for call control Loggers Controls logging to the console, system log, or log files Say Strings together audio files in various languages to provide feedback to say things like phone numbers, time of day, spell words, and so on Text-To-Speech (TTS) Interfaces with text-to-speech engines Timers POSIX or Linux kernel timing in applications XML Interfaces Uses XML for Call Detail Records (CDRs), RADIUS, CURL, LDAP, RPC, and/or SCGI The following image shows what the FreeSWITCH architecture looks like and how the modules orbit the core of FreeSWITCH: By combining the functionality of the various module interfaces, FreeSWITCH can be configured to connect IP phones, POTS lines, and IP-based telephone services. It can also translate audio formats and interfaces with a custom menu system, which you can create by yourself. You can even control a running FreeSWITCH server from another machine. Let's start by taking a closer look at a pair of important module types. Important modules – Endpoint and Dialplan Endpoint modules are critically important and add some of the key features that make FreeSWITCH the powerful platform it is today. The primary role of these modules is to take certain common communication technologies and normalize them into a common abstract entity which we refer to as a session. A session represents a connection between FreeSWITCH and a particular protocol. There are several Endpoint modules that come with FreeSWITCH, which implement several protocols such as SIP, H.323, Jingle (Google Talk), and some others. We will spend some time examining one of the more popular modules named mod_sofia. Sofia-SIP (http://sofia-sip.sourceforge.net) is an open source project sponsored by Nokia, which provides a programming interface for the Session Initiation Protocol (SIP). We use this library in FreeSWITCH in a module we call mod_sofia. This module registers to all the hooks in FreeSWITCH necessary to make an Endpoint module, and translates the native FreeSWITCH constructs into SIP constructs and back again. Configuration information is taken from the central FreeSWITCH configuration files, which allows mod_sofia to load user-defined preferences and connection details. This allows FreeSWITCH to accept registration from SIP phones and devices, register to other SIP Endpoints such as service providers, send notifications, and provide services to the phones such as voicemail. The SIP protocol is defined by a number of RFC (request for comment) documents. The primary RFC can be found at http://www.ietf.org/rfc/rfc3261.txt When a SIP call is established between FreeSWITCH and another SIP device, it will show up in FreeSWITCH as an active session. If the call is inbound, it can be transferred or bridged to interactive voice response (IVR) menus, hold music, or one or more extensions, though numerous other options are available. Let's examine a typical scenario where an SIP phone registered as extension 2000 dials extension 2001 with the hope of establishing a call. First, the SIP phone sends a call setup message to mod_sofia over the network (mod_sofia is listening for such messages). After receiving the message, mod_sofia in turn parses the relevant details and passes the call into the core state machine in FreeSWITCH. The state machine (in the FreeSWITCH core) then sends the call into the ROUTING state. The next step is to locate the Dialplan module based on the configuration data for the calling Endpoint. The default and most widely used Dialplan module is the XML Dialplan module. This module is designed to look up a list of instructions from the central XML registry within FreeSWITCH. The XML Dialplan module will parse a series of XML extension objects using regular expression pattern-matching. As we are trying to call 2001, we hope to find an XML extension testing the destination_number field for something that matches 2001 and routes accordingly. The Dialplan is not limited to matching only a single extension. The XML Dialplan module builds a sort of task list for the call. Each extension that matches it will have its actions added to the call's task list. Assuming FreeSWITCH finds at least one extension, the XML Dialplan will insert instructions into the session object with the information it needs to try and connect the call to 2001. Once these instructions are in place, the state of the calling session changes from ROUTING to EXECUTE, where the next handler drills down the list and executes the instructions obtained during the ROUTING state. This is where the application interface comes into the picture. Each instruction is added to the session in the form of an application name and a data argument that will be passed to that application. The one we will use in this example is the bridge application. The purpose of this application is to create another session with an outbound connection, then connect the two sessions for direct audio exchange. The argument we will supply to bridge will be user/2001, which is the easiest way to generate a call to extension 2001. A Dialplan entry for 2001 might look like this: <extension name="example"><condition field="destination_number"expression="^2001$"><action application="bridge" data="user/2001"/></condition></extension> The extension is named example, and it has a single condition to match. If the condition is matched, it has a single application to execute. In plain language, the mentioned extension could be expressed like this: If the caller dialed 2001, this establishes a connection between the calling party and the endpoint (that is, telephone) at 2001. Consider how this happens. Once we have inserted the instructions into the session, the session's state will change to EXECUTE, and the FreeSWITCH core will use the data collected to perform the desired action. First, the default execute state handler will parse the command to execute bridge on user/2001, then it will look up the bridge application and pass the user/2001 data in. This will cause the FreeSWITCH core to create a new outbound session of the desired type. User 2001 is also a SIP phone, so user/2001 will resolve into a SIP dial string, which will be passed to mod_sofia to ask it to create a new outbound session. If the setup for that new session is successful, there will now be two sessions in the FreeSWITCH core. The bridge application will take the new session and the original session (the caller's phone) and call the bridge function on it. This allows the audio to flow in both directions once the person at extension 2001 actually answers the phone. If that user was unable to answer or was busy, a timeout (that is, a failure) would occur and send the corresponding message back to the caller's phone. If a call is unanswered or an extension is busy, many routing options are possible, including call forwarding or voicemail. All of this happens from the simple action of picking up the phone handset and dialing 2 0 0 1. FreeSWITCH takes all of the complexity of SIP and reduces it to a common denominator. From there, it reduces the complexity further by allowing us to configure a single instruction in the Dialplan to connect the phone at 2000 to the phone at 2001. If we want to allow the phone at 2001 to be able to call the phone at 2000, we can add another entry in the Dialplan going the other way: <extension name="example 2"><condition field="destination_number" expression="^2000$"><action application="bridge" data="user/2000"/></condition></extension> In this scenario, the Endpoint module turned SIP into a FreeSWITCH session and the Dialplan module turned XML into an extension. The bridge application turned the complex code of creating an outbound call and connecting the audio into a simple application/data pair. Both the Dialplan module and the application module interface are designed around regular FreeSWITCH sessions. Therefore, not only does the abstraction make life easier for us at the user level, it also simplifies the design of the application and the Dialplan because they can be made agnostic of the actual endpoint technology involved in the call. It is because of this abstraction, when we make up a new Endpoint module tomorrow for something like Skype (there is actually such a thing present, by the way), that we can reuse all the same application and Dialplan modules. The same principle applies to the Say, Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and other such modules. It is possible that you may want to work with some specific data provided by the Endpoint's native protocol. In SIP, for instance, there are several arbitrary headers as well as several other bits of interesting data from the SIP packets. We solve this problem by adding variables to the channel. Using channel variables, mod_sofia can set these arbitrary values as they are encountered in the SIP data where you can retrieve them by name from the channel in your Dialplan or application. This way, we share our knowledge of these special variables with the SIP Endpoint. However, the FreeSWITCH core just sees them as arbitrary channel variables that the core can ignore. There are also several special reserved channel variables that can influence the behavior of FreeSWITCH in many interesting ways. If you have ever used a scripting language or configuration engine that uses variables (sometimes called attribute-value pairs or AVPs), you are at an advantage because channel variables are pretty much the same concept. There is simply a variable name and a value that is passed to the channel and the data is set. There is even an application interface for this, the set application, which lets you set your own variables from the Dialplan: <extension name="example 3"><condition field="destination_number" expression="^2000$"><action application="set" data="foo=bar"/><action application="bridge" data="user/2000"/></condition></extension> This example is almost identical to the previous example, but instead of just placing the call, we first set the variable foo equal to the value bar. This variable will remain set throughout the call and can even be referenced at the end of the call in the detail logs. The more we build things in small pieces, the more the same underlying resources can be reused, making the system simpler to use. For example, the codec interface knows nothing else about the core, other than its own isolated world of encoding and decoding audio packets. Once a proper codec module has been written, it becomes usable by any Endpoint interface capable of carrying that codec in its audio stream. This means that if we get a Text-To-Speech module working, we can generate synthesized speech on any and all Endpoints that FreeSWITCH supports. It does not matter which one comes first as they have nothing to do with each other. However, the addition of either one instantly adds functionality to the other. The TTS module becomes more useful because it can use more codecs; the codecs have become more useful because we added a new function that can take advantage of them. The same idea applies to applications. If we write a new application module, the existing endpoints will immediately be able to run and use that application.
Read more
  • 0
  • 0
  • 18983
article-image-openstack-networking-and-security-with-ansible-2
Vijin Boricha
03 Jul 2018
9 min read
Save for later

Automating OpenStack Networking and Security with Ansible 2 [Tutorial]

Vijin Boricha
03 Jul 2018
9 min read
OpenStack is a software that can help us build a system similar to popular cloud providers, such as AWS or GCP. OpenStack provides an API and a dashboard to manage the resources that it controls. Basic operations, such as creating and managing virtual machines, block storage, object storage, identity management, and so on, are supported out of the box. This is an excerpt from Ansible 2 Cloud Automation Cookbook written by Aditya Patawari, Vikas Aggarwal. No matter the Cloud Platform you are using this book will help you orchestrate your cloud infrastructure.  In the case of OpenStack, we control the underlying hardware and network, which comes with its own pros and cons. In this article we will leverage Ansible 2 to automate not so common networking tasks in OpenStack. We can use custom network solutions. We can use economical equipment or high-end devices, depending upon the actual need. This can help us get the features that we want and may end up saving money. Caution: Although OpenStack can be hosted on premises, several cloud providers provide OpenStack as a service. Sometimes these cloud providers may choose to turn off certain features or provide add-on features. Sometimes, even while configuring OpenStack on a self-hosted environment, we may choose to toggle certain features or configure a few things differently. Therefore, inconsistencies may occur. All the code examples in this article are tested on a self-hosted OpenStack released in August 2017, named Pike. The underlying operating system was CentOS 7.4. Managing security groups Security groups are the firewalls that can be used to allow or disallow the flow of traffic. They can be applied to virtual machines. Security groups and virtual machines have a many-to-many relationship. A single security group can be applied to multiple virtual machines and a single virtual machine can have multiple security groups. How to do it… Let's create a security group as follows: - name: create a security group for web servers os_security_group: name: web-sg state: present description: security group for web servers The name parameter has to be unique. The description parameter is optional, but we recommend using it to state the purpose of the security group. The preceding task will create a security group for us, but there are no rules attached to it. A firewall without any rules is of little use. So let's go ahead and add a rule to allow access to port 80 as follows: - name: allow port 80 for http os_security_group_rule: security_group: web-sg protocol: tcp port_range_min: 80 port_range_max: 80 remote_ip_prefix: 0.0.0.0/0 We also need SSH access to this server, so we should allow port 22 as well: - name: allow port 80 for SSH os_security_group_rule: security_group: web-sg protocol: tcp port_range_min: 22 port_range_max: 22 remote_ip_prefix: 0.0.0.0/0 How it works… For this module, we need to specify the name of the security group. The rule that we are creating will be associated with this group. We have to supply the protocol and the port range information. If we just want to whitelist only one port, then that would be the upper and lower bound of the range. Lastly, we have to specify the allowed addresses in the form of CIDR. The address 0.0.0.0/0 signifies that port 80 is open for everyone. This task will add an ingress type rule and allow traffic on port 80 to reach the instance. Similarly, we have to add a rule to allow traffic on port 22 as well. Managing network resources A network is a basic building block of the infrastructure. Most of the cloud providers will supply a sample or default network. While setting up a self-hosted OpenStack instance, a single network is typically created automatically. However, if the network is not created, or if we want to create another network for the purpose of isolation or compliance, we can do so using the os_network module. How to do it… Let's go ahead and create an isolated network and name it private, as follows: - name: creating a private network os_network: state: present name: private In the preceding example, we created a logical network with no subnets. A network with no subnets is of little use, so the next step would be to create a subnet: - name: creating a private subnet os_subnet: state: present network_name: private name: app cidr: 192.168.0.0/24 dns_nameservers: - 8.8.4.4 - 8.8.8.8 host_routes: - destination: 0.0.0.0/0 nexthop: 104.131.86.234 - destination: 192.168.0.0/24 nexthop: 192.168.0.1 How it works… The preceding task will create a subnet named app in the network called private. We have also supplied a CIDR for the subnet, 192.168.0.0/24. We are using Google DNS for nameservers as an example here, but this information should be obtained from the IT department of the organization. Similarly, we have set up the example host routes, but this information should be obtained from the IT department as well. After successful execution of this recipe, our network is ready to use. User management OpenStack provides an elaborate user management mechanism. If we are coming from a typical third-party cloud provider, such as AWS or GCP, then it can look overwhelming. The following list explains the building blocks of user management: Domain: This is a collection of projects and users that define an administrative entity. Typically, they can represent a company or a customer account. For a self-hosted setup, this could be done on the basis of departments or environments. A user with administrative privileges on the domain can further create projects, groups, and users. Group: A group is a collection of users owned by a domain. We can add and remove privileges from a group and our changes will be applied to all the users within the group. Project: A project creates a virtual isolation for resources and objects. This can be done to separate departments and environments as well. Role: This is a collection of privileges that can be applied to groups or users. User: A user can be a person or a virtual entity, such as a program, that accesses OpenStack services. For a complete documentation of the user management components, go through the OpenStack Identity document at https://docs.openstack.org/keystone/pike/admin/identity-concepts.html.  How to do it… Let's go ahead and start creating some of these basic building blocks of user management. We should note that, most likely, a default version of these building blocks will already be present in most of the setups: We'll start with a domain called demodomain, as follows: - name: creating a demo domain os_keystone_domain: name: demodomain description: Demo Domain state: present register: demo_domain After we get the domain, let's create a role, as follows: - name: creating a demo role os_keystone_role: state: present name: demorole Projects can be created, as follows: - name: creating a demo project os_project: state: present name: demoproject description: Demo Project domain_id: "{{ demo_domain.id }}" enabled: True Once we have a role and a project, we can create a group, as follows: - name: creating a demo group os_group: state: present name: demogroup description: "Demo Group" domain_id: "{{ demo_domain.id }}" Let's create our first user: - name: creating a demo user os_user: name: demouser password: secret-pass update_password: on_create email: demo@example.com domain: "{{ demo_domain.id }}" state: present Now we have a user and a group. Let's add the user to the group that we created before: - name: adding user to the group os_user_group: user: demouser group: demogroup We can also associate a user or a group with a role: - name: adding role to the group os_user_role: group: demo2 role: demorole domain: "{{ demo_domain.id }}" How it works… In step 1, the os_keystone_domain would take a name as a mandatory parameter. We also supplied a description for our convenience. We are going to use the details of this domain, so we saved it in a variable called demo_domain. In step 2, the os_keystone_role would just take a name and create a role. Note that a role is not tied up with a domain. In step 3, the os_project module would require a name. We have added the description for our convenience. The projects are tied to a domain, so we have used the demo_domain variable that we registered in a previous task. In step 4, the groups are tied to domains as well. So, along with the name, we would specify the description and domain ID like we did before. At this point, the group is empty, and there are no users associated with this group. In step 5, we supply name along with a password for the user. The update_password parameter is set to on_create, which means that the password won't be modified for an existing user. This is great for the sake of idempotency. We also specify the email ID. This would be required for recovering the password and several other use cases. Lastly, we add the domain ID to create the user in the right domain. In step 6, the os_user_group module will help us associate the demouser with the demogroup. In step 7, the os_user_role will take a parameter for user or group and associate it with a role. A lot of these divisions might not be required for every organization. We recommend going through the documentation and understanding the use case of each of them. Another point to note is that we might not even see the user management bits on a day-to-day basis. Depending on the setup and our responsibilities, we might only interact with modules that involve managing resources, such as virtual machines and storage. We learned how to successfully to solve complex OpenStack networking tasks with Ansible 2. To learn more about managing other public cloud platforms like AWS and Azure refer to our book  Ansible 2 Cloud Automation Cookbook. Getting Started with Ansible 2 System Architecture and Design of Ansible An In-depth Look at Ansible Plugins
Read more
  • 0
  • 0
  • 18977

article-image-new-functionality-opencv-30
Packt
25 Aug 2014
5 min read
Save for later

New functionality in OpenCV 3.0

Packt
25 Aug 2014
5 min read
In this article by Oscar Deniz Suarez, coauthor of the book OpenCV Essentials, we will cover the forthcoming Version 3.0, which represents a major evolution of the OpenCV library for Computer Vision. Currently, OpenCV already includes several new techniques that are not available in the latest official release (2.4.9). The new functionality can be already used by downloading and compiling the latest development version from the official repository. This article provides an overview of some of the new techniques implemented. Other numerous lower-level changes in the forthcoming Version 3.0 (updated module structure, C++ API changes, transparent API for GPU acceleration, and so on) are not discussed. (For more resources related to this topic, see here.) Line Segment Detector OpenCV users have had the Hough transform-based straight line detector available in the previous versions. An improved method called Line Segment Detector (LSD) is now available. LSD is based on the algorithm described at http://dx.doi.org/10.5201/ipol.2012.gjmr-lsd. This method has been shown to be more robust and faster than the best previous Hough-based detector (the Progressive Probabilistic Hough Transform). The detector is now part of the imgproc module. OpenCV provides a short sample code ([opencv_source_code]/samples/cpp/lsd_lines.cpp), which shows how to use the LineSegmentDetector class. The following table shows the main components of the class: Method Function <constructor> The constructor allows to enter parameters of the algorithm; particularly; the level of refinements we want in the result detect This method detects line segments in the image drawSegments This method draws the segments in a given image compareSegments This method draws two sets of segments in a given image. The two sets are drawn with blue and red color lines Connected components The previous versions of OpenCV have included functions for working with image contours. Contours are the external limits of connected components (that is, regions of connected pixels in a binary image). The new functions, connectedComponents and connectedComponentsWithStats retrieve connected components as such. The connected components are retrieved as a labeled image with the same dimensions as the input image. This allows drawing the components on the original image easily. The connectedComponentsWithStats function retrieves useful statistics about each component shown in the following table: CC_STAT_LEFT  The leftmost (x) coordinate, which is the inclusive start of the bounding box in the horizontal direction CC_STAT_TOP  The topmost (y) coordinate, which is the inclusive start of the bounding box in the vertical direction CC_STAT_WIDTH  The horizontal size of the bounding box CC_STAT_HEIGHT  The vertical size of the bounding box CC_STAT_AREA  The total area (in pixels) of the connected component Scene text detection Text recognition is a classic problem in Computer Vision. Thus, Optical Character Recognition (OCR) is now routinely used in our society. In OCR, the input image is expected to depict typewriter black text over white background. In the last years, researchers aim at the more challenging problem of recognizing text "in the wild" on street signs, indoor signs, with diverse backgrounds and fonts, colors, and so on. The following figure shows and example of the difference between the two scenarios. In this scenario, OCR cannot be applied to the input images. Consequently, text recognition is actually accomplished in two steps. The text is first localized in the image and then character or word recognition is performed on the cropped region. OpenCV now provides a scene text detector based on the algorithm described in Neumann L., Matas J.: Real-Time Scene Text Localization and Recognition, CVPR 2012 (Providence, Rhode Island, USA). The implementation of OpenCV makes use of additional improvements found at http://158.109.8.37/files/GoK2013.pdf. OpenCV includes an example ([opencv_source_code]/samples/cpp/textdetection.cpp) that detects and draws text regions in an input image. The KAZE and AKAZE features Several 2D features have been proposed in the computer vision literature. Generally, the two most important aspects in feature extraction algorithms are computational efficiency and robustness. One of the latest contenders is the KAZE (Japanese word meaning "Wind") and Accelerated-KAZE (AKAZE) detector. There are reports that show that KAZE features are both robust and efficient, as compared with other widely-known features (BRISK, FREAK, and so on). The underlying algorithm is described in KAZE Features, Pablo F. Alcantarilla, Adrien Bartoli, and Andrew J. Davison, in European Conference on Computer Vision (ECCV), Florence, Italy, October 2012. As with other keypoint detectors in OpenCV, the KAZE implementation allows retrieving both keypoints and descriptors (that is, a feature vector computed around the keypoint neighborhood). The detector follows the same framework used in OpenCV for other detectors, so drawing methods are also available. Computational photography One of the modules with most improvements in the forthcoming Version 3.0 is the computational photography module (photo). The new techniques include the functionalities mentioned in the following table: Functionality Description HDR imaging Functions for handling High-Dynamic Range images (tonemapping, exposure alignment, camera calibration with multiple exposures, and exposure fusion) Seamless cloning Functions for realistically inserting one image into other image with an arbitrary-shape region of interest. Non-photorealistic rendering This technique includes non-photorealistic filters (such as pencil-like drawing effect) and edge-preserving smoothing filters (those are similar to the bilateral filter). New modules Finally, we provide a list with the new modules in development for version 3.0: Module name Description videostab Global motion estimation, Fast Marching method softcascade Implements a stageless variant of the cascade detector, which is considered more accurate shape Shape matching and retrieval. Shape context descriptor and matching algorithm, Hausdorff distance and Thin-Plate Splines cuda<X> Several modules with CUDA-accelerated implementations of other functions in the library Summary In this article, we learned about the different functionalities in OpenCV 3.0 and its different components. Resources for Article: Further resources on this subject: Wrapping OpenCV [article] A quick start – OpenCV fundamentals [article] Linking OpenCV to an iOS project [article]
Read more
  • 0
  • 0
  • 18965
Modal Close icon
Modal Close icon