Predicting the Future of GANs

If you have completed all of the exercises in the chapters of this book, you have come a long way in your quest to learn and code Generative adversarial networks (GANs) for various real-world applications. GANs have the potential to cause disruption in a number of different industries. Scientists and researchers have developed various GANs that can be used to build commercial applications. Throughout this book, we have explored and implemented some of the most famous GAN architectures.

So, let's recap what we have learned thus far:

• We started with a gentle introduction to GANs, and learned various important concepts.
• We then explored a 3D-GAN, which is a type of GAN than can generate 3D images. We trained the 3D-GAN to generate 3D models of real-world objects such as an airplane or a table.
• In the third chapter, we explored conditional GANs...

Our predictions about the future of GANs

In my opinion, the future of GANs will be characterized by the following:

• Open acceptance of GANs and their applications by the research community.
• Impressive results—GANs have so far shown very impressive results on tasks that were difficult to perform using conventional methods. Transforming low-resolution images to high-resolution images, for example, was previously quite a challenging task and was generally carried out using CNNs. GAN architectures, such as SRGANs or pix2pix, have shown the potential of GANs for this application, while the StackGAN network has proved useful for text-to-image synthesis tasks. Nowadays, anyone can create an SRGAN network and train it on their own images.
• Advancements in deep learning techniques.
• GANs being used in commercial applications.
• Maturation of the training process of GANs.

...

Potential future applications of GANs

The future of GANs is bright! There are several areas in which I think it is likely that GANs will be used in the near future:

• Creating infographics from text
• Generating website designs
• Compressing data
• Drug discovery and development
• Generating text
• Generating music

Creating infographics from text

Designing infographics is a lengthy process. It takes hours of labor and requires specific skills. In marketing and social promotions, infographics work like a charm; they are the main ingredient of social media marketing. Sometimes, due to the lengthy process of creation, companies have to settle with a less effective strategy. AI and GANs can help designers in the creative process.

...

Exploring GANs

Other GAN architectures that you can explore include the following:

• BigGAN: LARGE SCALE GAN TRAINING FOR HIGH FIDELITY NATURAL IMAGE SYNTHESIS (https://arxiv.org/pdf/1809.11096.pdf)
• WaveGAN: Synthesizing Audio with Generative Adversarial Networks (https://arxiv.org/abs/1802.04208)
• BEGAN: BEGAN: Boundary Equilibrium Generative Adversarial Networks (https://arxiv.org/abs/1703.10717)
• AC-GAN: Conditional Image Synthesis With Auxiliary Classifier GANs (https://arxiv.org/abs/1610.09585)

• AdaGAN: AdaGAN: Boosting Generative Models (https://arxiv.org/abs/1701.02386v1)
• ArtGAN: ArtGAN: Artwork Synthesis with Conditional Categorial GANs (https://arxiv.org/abs/1702.03410)
• BAGAN: BAGAN: Data Augmentation with Balancing GAN (https://arxiv.org/abs/1803.09655)
• BicycleGAN: Toward Multimodal Image-to-Image Translation (https://arxiv.org/abs/1711.11586)
• CapsGAN: CapsGAN: Using...

Summary

In this book, my intention was to give you a taste of GANs and their applications in the world. The only limit is your imagination. There is an enormous list of different GAN architectures available, and they are becoming increasingly mature. GANs still have a fair way to go, because they still have problems, such as training instability and mode collapse, but various solutions have now been proposed, including label smoothing, instance normalization, and mini-batch discrimination. I hope that this book has helped you in the implementation of GANs for your own purposes. If you have any queries, drop me an email at [e-mail].