Research map / Detection and segmentation / CNN features
Region-aware Distribution Contrast: A Novel Approach to Multi-Task Partially Supervised Learning
Meixuan Li, Tianyu Li, Guoqing Wang, Peng Wang, Yang Yang, Jie Zou
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 |
|---|---|---|---|
| use SAM regions instead of patches for local contrast (ablation) | Gaussian contrast on HRNet18 · NYU-v2 one label | Normals mErr: 7.7141 -> 4.8613 | Improved |
| use SAM regions instead of patches for local contrast (ablation) | Gaussian contrast on HRNet18 · NYU-v2 one label | Semantic mIoU: 41.56 -> 42.28 | Improved |
| model regions as Gaussian distributions instead of vectors (ablation) | Region-aware contrast · NYU-v2 one label | Normals mErr: 5.8098 -> 4.8613 | Improved |
| model regions as Gaussian distributions instead of vectors (ablation) | Region-aware contrast · Cityscapes one label | Depth aErr: 0.0145 -> 0.0121 | Improved |
| Gaussian distribution instead of pixel contrast (ablation) | Region-aware contrast · NYU-v2 random label | Semantic mIoU: 46.41 -> 46.21 | Mixed |
| use Wasserstein distance instead of KL divergence (ablation) | Gaussian region contrast · NYU-v2 random label | Normals mErr: 5.04 -> 4.49 | Improved |
| use Wasserstein distance instead of KL divergence (ablation) | Gaussian region contrast · NYU-v2 one label | Semantic mIoU: 41.15 -> 42.28 | Improved |
| add region-wise Gaussian contrast to TaskExpert | TaskExpert ViT-Large · NYU-v2 one label | Semantic mIoU: 49.53 -> 55.81 | Improved |
| add region-wise Gaussian contrast in the fully supervised setting | MTL HRNet18 · NYU-v2 fully supervised | Normals mErr: 6.7298 -> 3.2878 | Improved |
| use SAM regions with vector contrast only (ablation) | Region-aware contrast · NYU-v2 fully supervised | Semantic mIoU: 47.20 -> 47.64 | No clear effect |
Closest papers by meaning
- Region Similarity Representation Learning (ICCV 2021 · similarity 0.88)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.88)
- RegionCL: Exploring Contrastive Region Pairs for Self-Supervised Representation Learning (ECCV 2022 · similarity 0.87)
- Region-based Cluster Discrimination for Visual Representation Learning (ICCV 2025 · similarity 0.87)
- 3D-Aware Multi-Task Learning with Cross-View Correlations for Dense Scene Understanding (CVPR 2026 · similarity 0.87)
- Dense Siamese Network for Dense Unsupervised Learning (ECCV 2022 · 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.
