Research map / Label-efficient and robust learning / Domain adaptation and generalization
360SFUDA++: Towards Source-Free UDA for Panoramic Segmentation by Learning Reliable Category Prototypes
Zheng Xu, Pengyuan Zhou, Athanasios V. Vasilakos, Lin Wang
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 reliable prototype adaptation (L_un + L_ppa) to supervised source loss (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 38.65 -> 51.05 | Improved |
| add L_un to supervised loss (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 38.65 -> 43.05 | Improved |
| add L_ppa on top of L_sup and L_un (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 43.05 -> 51.05 | Improved |
| add source-model fine-tuning loss L_sft (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 51.05 -> 51.16 | Improved |
| add cross-dual-attention module (CDAM) alone (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 38.65 -> 44.24 | Improved |
| add batch-norm statistics loss L_bns to CDAM (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 44.24 -> 44.79 | Improved |
| use RP2AM instead of PPAM (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 45.42 -> 51.05 | Improved |
| use cross-projection dual attention instead of within-feature dual attention (ablation) | CDAM · Cityscapes-to-DensePASS | mIoU: 41.53 -> 44.24 | Improved |
| use fixed FoV projection of 90 degrees versus none (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 38.65 -> 44.28 | Improved |
| enlarge FFP field of view from 90 to 360 degrees (ablation) | 360SFUDA++ · Cityscapes-to-DensePASS | mIoU: 44.28 -> 40.31 | Hurt |
| raise CDAM trade-off gamma from 0.1 to 0.2 (ablation) | CDAM · Cityscapes-to-DensePASS | mIoU: 44.24 -> 43.07 | Hurt |
| upgrade 360SFUDA to 360SFUDA++ with SegFormer-B2 | 360SFUDA · Cityscapes-to-DensePASS | mIoU: 50.12 -> 52.99 | Improved |
Closest papers by meaning
- Semantics Distortion and Style Matter: Towards Source-free UDA for Panoramic Segmentation (CVPR 2024 · similarity 0.97)
- GoodSAM: Bridging Domain and Capacity Gaps via Segment Anything Model for Distortion-aware Panoramic Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.88)
- Geometric Exploitation for Indoor Panoramic Semantic Segmentation (NeurIPS 2024 · similarity 0.87)
- AdapToPASS: Ambiguity-aware Adaptive Spherical Transformer for Panoramic Semantic Segmentation (NeurIPS 2026 · similarity 0.87)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · 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.
