Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unlocking Constraints: Source-Free Occlusion-Aware Seamless Segmentation
Yihong Cao, Jiaming Zhang, Xu Zheng, Hao Shi, Kunyu Peng, Hang Liu, Kailun Yang, Hui 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 OPLL (omni pseudo-labels with class-wise self-tuning thresholds and uncertainty-guided loss) (ablation) | UNLOCK · K2B | mAPQ: 22.13 -> 24.71 | Improved |
| use standard filtered pseudo-labels for self-training (ablation) | source model · K2B | mAAP: 10.22 -> 10.09 | Hurt |
| use all predictions directly as pseudo-labels (ablation) | source model · K2B | mAAP: 10.22 -> 10.15 | No clear effect |
| add ADCL spatial-aware amodal-driven mixing (ablation) | UNLOCK · K2B | mAPQ: 22.13 -> 25.84 | Improved |
| combine OPLL and ADCL (ablation) | UNLOCK · K2B | mAPQ: 22.13 -> 26.41 | Improved |
| replace zeroing only overlap regions with random values (ablation) | ADCL · K2B | mAPQ: 25.84 -> 24.81 | Hurt |
| zero out full amodal object regions (AoMix) instead of only overlaps (ablation) | ADCL · K2B | mAPQ: 25.84 -> 25.32 | Hurt |
| use instance masks instead of amodal instance masks for mixing (ablation) | ADCL · K2B | mAPQ: 25.71 -> 25.03 | Hurt |
| swap backbone among Trans4PASS, EDAPS, UnmaskFormer (ablation) | UNLOCK · K2B | mAPQ: 26.1 / 26.3 / 26.4 | No clear effect |
| apply UNLOCK instead of 360SFUDA++ | panoramic semantic segmentation · C2D | mIoU: 52.99 -> 54.55 | Improved |
Closest papers by meaning
- 360SFUDA++: Towards Source-Free UDA for Panoramic Segmentation by Learning Reliable Category Prototypes (TPAMI 2024 · similarity 0.90)
- Semantics Distortion and Style Matter: Towards Source-free UDA for Panoramic Segmentation (CVPR 2024 · similarity 0.89)
- Occlusion-Aware Seamless Segmentation (ECCV 2024 · similarity 0.89)
- GoodSAM: Bridging Domain and Capacity Gaps via Segment Anything Model for Distortion-aware Panoramic Semantic Segmentation (CVPR 2024 · similarity 0.88)
- More than the Sum: Panorama-Language Models for Adverse Omni-Scenes (CVPR 2026 · similarity 0.85)
- AdapToPASS: Ambiguity-aware Adaptive Spherical Transformer for Panoramic Semantic Segmentation (NeurIPS 2026 · similarity 0.84)
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.
