Research map / Image generation / Editing and style transfer
Identity From Here, Pose From There: Self-Supervised Disentanglement and Generation of Objects Using Unlabeled Videos
Fanyi Xiao, Haotian Liu, Yong Jae Lee
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 pixel verification loss (ablation) | identity-pose generator · YTBB car | FID: 18.47 -> 18.03 | Improved |
| add pixel verification loss (ablation) | identity-pose generator · YTBB bus | FID: 19.73 -> 16.77 | Improved |
| add pixel verification loss (ablation) | identity-pose generator · YTBB truck | FID: 26.33 -> 24.87 | Improved |
| add L2dis reconstruction terms to the disentanglement loss (ablation) | generator without Lpv · YTBB truck | FID: 37.38 -> 26.33 | Improved |
| add L2dis reconstruction terms (ablation) | generator without Lpv · YTBB bus | FID: 24.88 -> 19.73 | Improved |
| add L2dis reconstruction terms (ablation) | generator without Lpv · YTBB car | LPIPS: 0.35 -> 0.34 | Improved |
| add explicit disentanglement loss and Lpv on top of direct pseudo-target supervision | Pix2PixHD baseline · YTBB truck | FID: 75.82 -> 24.87 | Improved |
| add explicit disentanglement and Lpv losses over Pix2PixHD | Pix2PixHD baseline · YTBB car | FID: 25.18 -> 18.03 | Improved |
| use direct pseudo ground-truth supervision instead of cyclic constraints only | FusionImg · YTBB car | FID: 239.37 -> 18.03 | Improved |
| use direct pseudo ground-truth supervision instead of identity-confusion classifier | DrNet · YTBB bus | FID: 38.63 -> 16.77 | Improved |
| our method vs Pix2PixHD on simple synthetic data | generator · 3D Cars | FID: 97.76 -> 71.33 | Improved |
| our method vs Pix2PixHD on simple synthetic data | generator · 3D Chairs | FID: 31.01 -> 29.58 | No clear effect |
Closest papers by meaning
- Unsupervised Robust Disentangling of Latent Characteristics for Image Synthesis (ICCV 2019 · similarity 0.89)
- Collecting The Puzzle Pieces: Disentangled Self-Driven Human Pose Transfer by Permuting Textures (ICCV 2023 · similarity 0.87)
- Beyond Pixel Mimicry: Disentangled Self-Similarity Rewards for Diverse Subject-Driven Generation (ECCV 2026 · similarity 0.87)
- Realistic Face Reenactment via Self-Supervised Disentangling of Identity and Pose (AAAI 2020 · similarity 0.87)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · similarity 0.86)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.85)
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.
