VITA-Group/TransGAN
A generative adversarial network built entirely with Transformer encoders that scales to produce competitive image synthesis.
Not currently ranked — collecting fresh signals.
star history
TransGAN replaces convolutional layers in traditional GANs with pure Transformer architectures. The model generates images through an adversarial training process where a generator creates samples and a discriminator evaluates them, both implemented using Transformer encoders. It supports distributed training, gradient accumulation, and includes IS/FID evaluation metrics for benchmarking generated image quality.
Frequently asked
- What is VITA-Group/TransGAN?
- A generative adversarial network built entirely with Transformer encoders that scales to produce competitive image synthesis.
- Is TransGAN open source?
- Yes — VITA-Group/TransGAN is an open-source project tracked on heatdrop.
- What language is TransGAN written in?
- VITA-Group/TransGAN is primarily written in Python.
- How popular is TransGAN?
- VITA-Group/TransGAN has 1.7k stars on GitHub.
- Where can I find TransGAN?
- VITA-Group/TransGAN is on GitHub at https://github.com/VITA-Group/TransGAN.