Research map / Image generation / Editing and style transfer
SPatchGAN: A Statistical Feature Based Discriminator for Unsupervised Image-to-Image Translation
Xuning Shao, Weidong Zhang
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 channel-wise mean feature (ablation) | SPatchGAN discriminator · Selfie-to-Anime | FID: 84.1 -> 83.3 | Improved |
| add channel-wise mean feature (ablation) | SPatchGAN discriminator · Selfie-to-Anime | KID: 0.0228 -> 0.0214 | Improved |
| add channel-wise max feature (ablation) | SPatchGAN discriminator · Selfie-to-Anime | FID: 83.8 -> 83.3 | Improved |
| add channel-wise standard deviation feature (ablation) | SPatchGAN discriminator · Selfie-to-Anime | FID: 84.9 -> 83.3 | Improved |
| add channel-wise standard deviation feature (ablation) | SPatchGAN discriminator · Selfie-to-Anime | KID: 0.0214 -> 0.0214 | No clear effect |
| use SPatchGAN discriminator instead of multi-scale PatchGAN (ablation) | SPatchGAN · Selfie-to-Anime | FID: 94.0 -> 83.3 | Improved |
| use SPatchGAN discriminator instead of multi-scale PatchGAN (ablation) | SPatchGAN · Selfie-to-Anime | KID: 0.0362 -> 0.0214 | Improved |
| add weak cycle constraint (ablation) | SPatchGAN · Selfie-to-Anime | FID: 82.4 -> 83.3 | Hurt |
| add weak cycle constraint (ablation) | SPatchGAN · Male-to-Female | FID: 11.4 -> 8.73 | Improved |
| add weak cycle constraint (ablation) | SPatchGAN · Glasses Removal | FID: 16.2 -> 13.9 | Improved |
| replace CycleGAN discriminator with SPatchGAN discriminator (ablation) | CycleGAN · Male-to-Female | FID: 24.5 -> 13.4 | Improved |
| replace CycleGAN discriminator with SPatchGAN discriminator (ablation) | CycleGAN · Male-to-Female | KID: 0.0240 -> 0.0107 | Improved |
Later papers that compared with this method
1 later accepted paper reports SpatchGAN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Unsupervised Image-to-Image Translation with Density Changing Regularization (NeurIPS 2022 · better on 2 of 2 shared results)
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.
- U-GAT-IT — U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation (ICLR 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
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.87)
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · similarity 0.87)
- Unpaired Image-to-Image Translation using Adversarial Consistency Loss (ECCV 2020 · similarity 0.87)
- StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018 · similarity 0.86)
- Few-Shot Unsupervised Image-to-Image Translation (ICCV 2019 · similarity 0.86)
- Unaligned Image-to-Image Translation by Learning to Reweight (ICCV 2021 · similarity 0.86)
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.
