Research map / Detection and segmentation / Open-set and few-shot detection
Human De-Occlusion: Invisible Perception and Recovery for Humans
Qiang Zhou, Shiyin Wang, Yitong Wang, Zilong Huang, Xinggang Wang
Check 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 |
|---|---|---|---|
| Use the full two-stage human de-occlusion method instead of PCNets | AHP mask completion · AHP Syn. and AHP Real | amodal mask ℓ1: 0.1959 -> 0.1500; 0.2218 -> 0.1635 | Improved |
| Use the full two-stage human de-occlusion method instead of PCNets | AHP mask completion · AHP Syn. and AHP Real | IoU (amodal/invisible): 83.1/29.1 -> 84.6/43.7; 81.3/31.2 -> 86.1/40.3 | Improved |
| Use the full recovery method instead of PCNets | AHP content recovery · AHP Syn. and AHP Real | FID: 18.50 -> 13.85; 28.30 -> 19.49 | Improved |
| Add a discriminator to mask completion (ablation) | Mask completion baseline · AHP fixed validation | amodal IoU: 84.0 -> 84.3 | Improved |
| Add modal segmentation (ablation) | Mask completion baseline · AHP fixed validation | amodal IoU: 84.3 -> 83.1 | Hurt |
| Add modal segmentation (ablation) | Mask completion baseline · AHP fixed validation | invisible IoU: 30.3 -> 34.6 | Improved |
| Add parsing branches (ablation) | Mask completion with discriminator and modal segmentation · AHP fixed validation | invisible IoU: 34.6 -> 40.0 | Improved |
| Add template pose masks (ablation) | Mask completion with discriminator and modal segmentation · AHP fixed validation | invisible IoU: 34.6 -> 38.5 | Improved |
| Use background proportion w=0.3 instead of w=1.0 (ablation) | Unet content recovery · AHP content recovery validation | FID: 19.71 -> 17.85 | Improved |
| Add body-part attention (ablation) | Parsing-guided content recovery · AHP content recovery validation | FID: 19.66 -> 16.76 | Improved |
| Add visible-to-invisible context attention (ablation) | Parsing-guided content recovery · AHP content recovery validation | FID: 19.66 -> 17.37 | Improved |
| Fuse both PGA attention streams rather than cascading them (ablation) | Parsing-guided content recovery · AHP content recovery validation | FID: 14.66 -> 13.85 | Improved |
Closest papers by meaning
- Amodal Scene Analysis via Holistic Occlusion Relation Inference and Generative Mask Completion (AAAI 2024 · similarity 0.88)
- Large Occluded Human Image Completion via Image-Prior Cooperating (AAAI 2024 · similarity 0.86)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.86)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · similarity 0.85)
- DPMesh: Exploiting Diffusion Prior for Occluded Human Mesh Recovery (CVPR 2024 · similarity 0.85)
- BLADE: Box-Level Supervised Amodal Segmentation through Directed Expansion (AAAI 2024 · 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.
