Why is it important to study GANs?

By Chamodi Jayathilaka

The world has grown intrigued about Artificial Intelligence and Machine Learning since the widespread release of ChatGPT. ChatGPT extends the GANs concept to text-based communication. It converses with humans in a natural manner by using GANs to produce responses to input text. This blog will explain why it is important to study GANs.

What are GANs?

Generative Adversarial Networks (GANs) are a subset of Machine Learning in which we feed the training dataset to the model and the model learns to generate new data with the same features as the training set it was fed. GANs are a type of Deep Generative Model that is much more powerful in generating new images. GANs were first presented by Goodfellow et al. in 2014 and the proposed architecture consists of two networks: 1) A generative network, called Generator (G) and 2) A discriminative network called a Discriminator (D). The generator can learn the underlying data distribution in training data while the discriminator can classify the output as a real sample or fake sample. The generator tries to generate fake data that look realistic, to fool the discriminator, while the discriminator tries to identify the real data (training data) from generated fake data.

The market for Generative Adversarial Networks (GANs) applications has seen a significant rise in the last couple of years, which has made this technology successful for high-fidelity natural image synthesis, text-to-image translation, music generation, video generation, 3D generation, and more practical applications. Let’s see some applications of GANs.

Applications of GANs

Most of the applications in this industry are mainly lying on computer vision. The common applications of GANs are:

  • Creation of music, realistic pictures, 2D and 3D objects, and human faces for anime.

  • Recognizing instances of fraud when hackers conduct an aggressive attack to get data.
  • The identification of tumors in human bodies by comparing picture datasets with those of organs in good health.
  • Media translations such as text-to-image and image-to-image translations.

  • Using super-resolution, photo mixing, and enhancing the existing image data in photographs.
  • Transforming images of people into Emojis or using filters from Instagram or Faceapp, etc.

Reasons for the excitement over GANs

  • GANs don’t need labeled data because they tend to learn effectively without supervision, which makes them extremely powerful and easy to understand by doing away with the laborious effort of labeling and annotating the data.
  • The Generative model may produce natural images of high quality that gradually improve to produce data that seems more and more real by coupling with an adversarial network.
  • Use adversarial networks to generate more data instead of using tricks like data augmentation.


GANs are an exciting and rapidly changing field, delivering on the promise of generative models in their ability to get real examples across a variety of problem domains. It will become a future tool one day. Despite all the difficulties presented by the past ten years of research, GANs have developed content that will become harder and harder to tell apart from actual content as time goes on. Because we’re currently discussing “what we can do for GANs” to make them more stable, we don’t yet know “what GANs can do for us”, but the future of GANs for humanity looks promising.


  • https://arxiv.org/abs/1406.2661
  • https://machinelearningmastery.com/what-are-generative-adversarial-networks-gans/
  • https://machinelearningmastery.com/impressive-applications-of-generative-adversarial-networks/