junyanz/BicycleGAN

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1514 stars · 252 forks observed · 2026-08-28

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

A PyTorch implementation of BicycleGAN, a model for multimodal image-to-image translation that generates diverse outputs from a single input by sampling a latent code. It is the official research code accompanying the NIPS 2017 paper by UC Berkeley and Adobe Research.

Use cases

• generate diverse day images from a night photo
• translate edges to realistic shoes with multiple styles
• convert labels to facades with varied appearances
• create morphing videos between translated images
• experiment with conditional GAN training on paired image data

When to choose

• you need multimodal image-to-image translation with diverse outputs
• you want a reference implementation of the BicycleGAN paper
• you have paired training data and a CUDA GPU

When to avoid

• you need unpaired image translation (use CycleGAN instead)
• you require actively maintained code with recent PyTorch versions
• you need production-ready inference at scale

Facets

library · maturity maintenance

deep-learning image-processing machine-learning deep-learning computer-vision image-processing machine-learning python gan pytorch image-to-image-translation generative-models research-code nips-2017 linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="junyanz/BicycleGAN")