Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly Supervised Learning of Semantic Correspondence through Cascaded Online Correspondence Refinement
Yiwen Huang, Yixuan Sun, Chenghang Lai, Qing Xu, Xiaomei Wang, Xuli Shen, Weifeng Ge
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 |
|---|---|---|---|
| pretrain matching models on SC-ImageNet pseudo-labels, then fine-tune | VAT · SPair-71k | PCK@0.10: 55.5 -> 60.3 | Improved |
| pretrain on SC-ImageNet then fine-tune | CATs · SPair-71k | PCK@0.10: 49.9 -> 54.6 | Improved |
| pretrain on SC-ImageNet then fine-tune | SCorrSAN · SPair-71k | PCK@0.10: 55.3 -> 58.9 | Improved |
| pretrain on SC-ImageNet then fine-tune | TransforMatcher · PF-WILLOW | PCK@0.05 / PCK@0.10: 48.9 -> 49.9 at 0.05; 76.0 -> 77.2 at 0.10 | Mixed |
| pretrain on SC-ImageNet then fine-tune | DHPF · PF-PASCAL | PCK@0.10: 90.7 -> 90.7 | No clear effect |
| add saliency detection to pseudo-label generator (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 55.9 -> 56.8 | Improved |
| add regularized Hough matching (RHM) (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 55.6 -> 56.8 | Improved |
| use independent (non-shared) gated cross-attention in each stage (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 52.6 -> 56.8 | Improved |
| include the region-level stage (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 51.0 -> 56.1 | Improved |
| include the pixel-level stage (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 56.1 -> 56.8 | Improved |
| raise input resolution from 256 to 512 (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 53.2 -> 56.8 | Improved |
| use RHM hyperpixel length 5 instead of 3 or 7 (ablation) | pseudo-label generator · PF-PASCAL | PCK@0.05: 56.0 (3) / 56.8 (5) / 56.2 (7) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TransforMatcher — TransforMatcher: Match-to-Match Attention for Semantic Correspondence (CVPR 2022) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Semi-Supervised Learning of Semantic Correspondence With Pseudo-Labels (CVPR 2022 · similarity 0.89)
- Learning Contrastive Representation for Semantic Correspondence (IJCV 2022 · similarity 0.88)
- Probabilistic Warp Consistency for Weakly-Supervised Semantic Correspondences (CVPR 2022 · similarity 0.88)
- Pyramidal Semantic Correspondence Networks (TPAMI 2021 · similarity 0.88)
- End-to-End Weakly-Supervised Semantic Alignment (CVPR 2018 · similarity 0.88)
- Weakly Supervised Manifold Learning for Dense Semantic Object Correspondence (ICCV 2017 · similarity 0.88)
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.
