Research map / Detection and segmentation / CNN features
TransforMatcher: Match-to-Match Attention for Semantic Correspondence
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 TransforMatcher with global match-to-match attention and multi-level correlation features | Semantic correspondence · SPair-71k test, α_bbox 0.1 (F) | PCK: 50.4 -> 53.7 | Helped |
| Add data augmentation (ablation) | TransforMatcher with rotary positional embedding · SPair-71k test, α_bbox 0.1 (F) | PCK: 50.2 -> 53.7 | Helped |
| Add data augmentation (ablation) | TransforMatcher with rotary positional embedding · PF-PASCAL test, α_img 0.1 (F) | PCK: 90.4 -> 91.8 | Helped |
| Replace absolute positional embedding with rotary positional embedding (ablation) | TransforMatcher without augmentation · SPair-71k test, α_bbox 0.1 (F) | PCK: 48.7 -> 50.2 | Helped |
| Replace absolute positional embedding with rotary positional embedding (ablation) | TransforMatcher with augmentation · SPair-71k test, α_bbox 0.1 (F) | PCK: 48.9 -> 53.7 | Helped |
| Use multi-channel correlation features instead of a single channel formed from concatenated features (ablation) | TransforMatcher · SPair-71k test, α_bbox 0.1 | PCK: 41.7 -> 53.7 | Helped |
| Use multi-channel correlation features instead of a single-channel mean correlation (ablation) | TransforMatcher · SPair-71k test, α_bbox 0.1 | PCK: 45.1 -> 53.7 | Helped |
| Use additive attention instead of Performer (ablation) | TransforMatcher attention module · SPair-71k test, α_bbox 0.1 | PCK: 48.8 -> 48.9 | No clear effect |
| Replace additive attention with Linformer (ablation) | TransforMatcher attention module · SPair-71k test, α_bbox 0.1 | PCK: 48.9 -> 1.3 | Hurt |
| Apply data augmentation (ablation) | TransforMatcher · SPair-71k class-wise PCK | PCK | Mixed |
| Transfer the PF-PASCAL-trained model to PF-WILLOW | TransforMatcher · PF-WILLOW test | PCK: 74.1 -> 66.7 | Hurt |
Closest papers by meaning
- MATCHA: Towards Matching Anything (CVPR 2025 · similarity 0.87)
- Unifying Feature and Cost Aggregation with Transformers for Semantic and Visual Correspondence (ICLR 2024 · similarity 0.87)
- SD4Match: Learning to Prompt Stable Diffusion Model for Semantic Matching (CVPR 2024 · similarity 0.87)
- Pixel-level Semantic Correspondence through Layout-aware Representation Learning and Multi-scale Matching Integration (CVPR 2024 · similarity 0.87)
- ETO:Efficient Transformer-based Local Feature Matching by Organizing Multiple Homography Hypotheses (NeurIPS 2024 · similarity 0.87)
- Jamais Vu: Exposing the Generalization Gap in Supervised Semantic Correspondence (NeurIPS 2025 · 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.