🎆🌆 Edge#231: Text-to-Image Synthesis with GANs

In this issue:

• we explore Text-to-image synthesis with GANs; 

• we discuss Google’s XMC-GAN, a modern approach to text-to-image synthesis; 

• we explore NVIDIA GauGAN2 Demo. 

Enjoy the learning!  

---

💡 ML Concept of the Day: Text-to-Image Synthesis with GANs 

In Edge#229, we discussed the VQGAN+CLIP method that leverages a pretrained model with a generative adversarial network (GAN) to create high-fidelity images based on textual input. In that model, CLIP learns the similarities between textual inputs and images to guide the GAN model. VQGAN+CLIP represents one of the newest methods to use GANs for text-to-image synthesis but certainly not the only one. As a matter of fact, GANs represent the most traditional text-to-image method in deep learning architectures.