What it does This repository is a hand-maintained bibliography of image-to-image translation research, organized by supervision level (supervised vs. unsupervised) and technique. Each entry lists the paper, venue, arXiv link, and — crucially — a code repository when one is available. The author sorts by first arXiv submission date, which keeps the timeline of ideas visible.

The interesting bit The real value is the taxonomy. The README splits unsupervised work into sub-genres — general methods, attention/instance-guided, many-to-many attribute translation, and disentangled/exemplar-guided — which helps you find the right flavor of GAN for your problem without reading fifty abstracts.

Key highlights

• Coverage from foundational pix2pix (2016) through few-shot and zero-shot variants (2019)
• Explicit code links for most entries, including unofficial reimplementations when the original authors didn’t release
• Video-to-video work included (vid2vid, Recycle-GAN, Everybody Dance Now)
• Open to PRs and issues — the author notes it is “constantly updating”
• No language dependencies; pure reference material

Caveats

• README is truncated in the source view, so full coverage of 2020+ work is unclear
• Some entries lack code links entirely (e.g., Contextual GAN, C2-GAN)
• “To be classified” section exists, suggesting the taxonomy is incomplete

Verdict Grab this if you’re entering the field or need to find a baseline method fast. Skip it if you want narrative synthesis or benchmark comparisons — this is a card catalog, not a review paper.