SummitKwan/transparent_latent_gan

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1973 stars · 356 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

TL-GAN is a Python/TensorFlow project that makes a GAN's latent space transparent by discovering feature axes, enabling controlled image synthesis and editing. It uses a coupled feature extractor network to correlate latent vectors with image features without retraining the GAN.

Use cases

• control features of GAN-generated images
• edit generated faces along attributes like age or gender
• add controllable knobs to a pretrained GAN without retraining
• explore and interpret a GAN's latent space
• generate photo-realistic images with specified traits

When to choose

• you have a pretrained GAN and want feature-controlled generation quickly
• you want to add new attribute controllers without retraining the generator
• you want an interpretable latent space for image editing experiments

When to avoid

• you need a maintained, production-ready image generation pipeline
• you want state-of-the-art GAN architectures like StyleGAN with native controls
• you need non-Python or non-TensorFlow 1.x environments

Facets

library · maturity maintenance

machine-learning deep-learning image-processing computer-vision deep-learning machine-learning image-processing computer-vision python gan latent-space controlled-generation tensorflow keras image-synthesis research gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="SummitKwan/transparent_latent_gan")