Research map / Detection and segmentation / CNN features
Multi-Scale Matching Networks for Semantic Correspondence
Dongyang Zhao, Ziyang Song, Zhenghao Ji, Gangming Zhao, Weifeng Ge, Yizhou Yu
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 local self attention (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 79.9 -> 81.1 | Improved |
| add dense connections across layers of a conv group (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 74.9 -> 81.1 | Improved |
| add cross-scale feature fusion (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 78.5 -> 81.1 | Improved |
| add complementary matching learning (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 80.0 -> 81.1 | Improved |
| add rich multi-scale supervision (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 31.4 -> 81.1 | Improved |
| add last layer (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.05: 77.9 -> 81.1 | Improved |
| add local self attention (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.1: 90.7 -> 91.6 | Improved |
| add cross-scale feature fusion (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.15: 94.3 -> 95.9 | Improved |
| add last layer (ablation) | MMNet · PF-PASCAL (ResNet-101-FCN) | PCK@0.1: 90.9 -> 91.6 | Mixed |
| change backbone from ResNet-101-FCN to VGG-16 | MMNet · PF-PASCAL | PCK@0.05: 81.1 -> 69.5 | Hurt |
| change backbone from ResNet-101-FCN to DeepLabV3-ResNet101 | MMNet · PF-PASCAL | PCK@0.05: 81.1 -> 73.3 | Hurt |
| change backbone from ResNet-50 to ResNet-101 | MMNet · PF-PASCAL | PCK@0.05: 75.3 -> 77.6 | Improved |
Closest papers by meaning
- Learning Contrastive Representation for Semantic Correspondence (IJCV 2022 · similarity 0.91)
- Deep Semantic Feature Matching (CVPR 2017 · similarity 0.90)
- Neural Matching Fields: Implicit Representation of Matching Fields for Visual Correspondence (NeurIPS 2022 · similarity 0.89)
- End-to-End Weakly-Supervised Semantic Alignment (CVPR 2018 · similarity 0.88)
- Dynamic Context Correspondence Network for Semantic Alignment (ICCV 2019 · similarity 0.88)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (CVPR 2024 · 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.
