Research map / Image generation / Editing and style transfer
Bi-Level Feature Alignment for Versatile Image Translation and Manipulation
Fangneng Zhan, Yingchen Yu, Rongliang Wu, Jiahui Zhang, Kaiwen Cui, Aoran Xiao, Shijian Lu, Chunyan Miao
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 |
|---|---|---|---|
| replace CoCosNet with RABIT | CoCosNet · ADE20K translation | FID: 26.40 -> 24.35 | Improved |
| replace CoCosNet with RABIT | CoCosNet · CelebA-HQ (Edge) translation | FID: 14.30 -> 11.67 | Improved |
| replace CoCosNet with RABIT | CoCosNet · DeepFashion translation | FID: 14.40 -> 12.58 | Improved |
| replace CoCosNet with RABIT | CoCosNet · CelebA-HQ (Semantic) translation | FID: 21.83 -> 20.44 | Improved |
| replace CoCosNet with RABIT | CoCosNet · ADE20K translation | LPIPS: 0.580 -> 0.571 | No clear effect |
| replace CoCosNet with RABIT | CoCosNet · ADE20K manipulation | FID: 41.03 -> 26.61 | Improved |
| replace CoCosNet with RABIT | CoCosNet · CelebA-HQ manipulation | FID: 68.70 -> 60.87 | Improved |
| replace CoCosNet with RABIT | CoCosNet · CelebA-HQ manipulation | LPIPS: 0.224 -> 0.176 | Improved |
| replace CoCosNet with RABIT | CoCosNet · ADE20K manipulation | PSNR: 20.30 -> 23.08 | Improved |
| add bi-level ranking and attention to build correspondence at 128x128 instead of lower resolution | CoCosNet · image translation | FID and SWD | Improved |
| use semantic position encoding instead of vanilla position encoding (ablation) | RABIT · warped exemplars | warping quality | Improved |
| add confidence feature injection (CMAP and Multi-CMAP) (ablation) | RABIT · image translation | image quality | 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.
- SEAN — SEAN: Image Synthesis With Semantic Region-Adaptive Normalization (CVPR 2020) · better on 3 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
- Unbalanced Feature Transport for Exemplar-Based Image Translation (CVPR 2021 · similarity 0.89)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.89)
- DA-GAN: Instance-Level Image Translation by Deep Attention Generative Adversarial Networks (CVPR 2018 · similarity 0.88)
- Generative Semantic Manipulation with Mask-Contrasting GAN (ECCV 2018 · similarity 0.88)
- Navigating the GAN Parameter Space for Semantic Image Editing (CVPR 2021 · similarity 0.88)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 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.
