Research map / Image generation / Editing and style transfer
Contrastive Learning for Unpaired Image-to-Image Translation
Taesung Park Alexei A. Efros Richard Zhang Jun-Yan Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add weight sharing between encoder and MLP for input and output patches (ablation) | CUT · Horse->Zebra | FID: 50.5 -> 46.6 | Improved |
| add PatchNCE gradient to the decoder (ablation) | CUT · Horse->Zebra | FID: 444.2 -> 46.6 | Improved |
| use CUT instead of CycleGAN | CUT · Cityscapes | FID: 76.3 -> 56.4 | Improved |
| use CUT instead of CycleGAN | CUT · Horse->Zebra | FID: 77.2 -> 45.5 | Improved |
| use CUT instead of CycleGAN | CUT · Cat->Dog | FID: 85.9 -> 76.2 | Improved |
| use CUT instead of CycleGAN | CUT · Cityscapes | mAP: 20.4 -> 24.7 | Improved |
| use FastCUT instead of CycleGAN | FastCUT · Horse->Zebra | FID: 77.2 -> 73.4 | Improved |
| use FastCUT instead of CycleGAN | FastCUT · Cat->Dog | FID: 85.9 -> 94.0 | Hurt |
| use CUT (one-sided) instead of CycleGAN, training time | CUT · Horse->Zebra | training speed: 0.40 -> 0.24 sec/iter | Improved |
| add identity loss PatchNCE on domain Y (ablation) | CUT · Horse->Zebra / Cityscapes | FID | Mixed |
| sample negatives from within the input image instead of external or added external negatives (ablation) | CUT · Cityscapes / Horse->Zebra | FID | Improved |
| use multiple encoder layers instead of last layer only (ablation) | CUT · Cityscapes / Horse->Zebra | FID | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DRIT — DRIT++: Diverse Image-to-Image Translation via Disentangled Representations (IJCV 2020) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Instance-Wise Hard Negative Example Generation for Contrastive Learning in Unpaired Image-to-Image Translation (ICCV 2021 · similarity 0.91)
- Exploring Patch-Wise Semantic Relation for Contrastive Learning in Image-to-Image Translation Tasks (CVPR 2022 · similarity 0.90)
- Robust Contrastive Learning Using Negative Samples with Diminished Semantics (NeurIPS 2021 · similarity 0.90)
- Unpaired Image-To-Image Translation Using Cycle-Consistent Adversarial Networks (ICCV 2017 · similarity 0.89)
- Patch-Wise Graph Contrastive Learning for Image Translation (AAAI 2024 · similarity 0.87)
- Rethinking the Paradigm of Content Constraints in Unpaired Image-to-Image Translation (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
