It is important to understand the different kinds of object recognition tasks, as the required neural network architecture greatly depends on the task. Object recognition tasks can be broadly classified into three different types:
- Image classification
- Object detection
- Instance segmentation
The following diagram depicts the difference between each task:
![](https://static.packt-cdn.com/products/9781789138900/graphics/assets/ce20900f-6812-4db0-863f-02f28952210f.png)
In Image Classification, the input to the problem is an image and the required output is simply a prediction of the class that the image belongs to. This is analogous to our first project, where we constructed a classifier to predict whether a patient is at risk of diabetes. In image classification, the problem is applied on pixels as our input data (specifically, the intensity value of each pixel), instead of tabular data represented by pandas DataFrames. In this project, we will focus on image...