Research map / Detection and segmentation / CNN features
ComPtr: Toward Diverse Bi-Source Dense Prediction Tasks via a Simple Yet General Complementary Transformer
Youwei Pang, Xiaoqi Zhao, Lihe Zhang, Huchuan Lu
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 difference awareness block (DAB) (ablation) | ComPtr without CEB · LEVIR-CD | IoU: 84.30 -> 86.25 | Improved |
| add consistency enhancement block (CEB) (ablation) | ComPtr without DAB · LEVIR-CD | IoU: 85.41 -> 86.25 | Improved |
| add both CEB and DAB (ablation) | baseline · RGBT-CC | GAME0: 12.53 -> 10.52 | Improved |
| add complementarity operations (CompOps) (ablation) | pure cross-source transformer · LEVIR-CD | IoU: 84.81 -> 86.25 | Improved |
| place CEB in encoder and DAB in decoder rather than swapped (ablation) | ComPtr · LEVIR-CD | IoU: 82.65 -> 86.25 | Improved |
| use LAA structure in CompOps rather than LAAA (ablation) | ComPtr · LEVIR-CD | IoU: 85.21 -> 86.25 | Improved |
| use K=4 prototypes in ADA instead of standard attention (ablation) | ComPtr · RGBT-CC | GAME0: 11.11 -> 10.52 | Improved |
| use K=4 rather than K=1 prototypes (ablation) | ADA · RGBT-CC | GAME0: 11.59 -> 10.52 | Improved |
| use K=4 rather than K=infinity prototypes (ablation) | ADA · LEVIR-CD | IoU: 86.01 -> 86.25 | Improved |
| deepen decoder from one DEB per stage to 4,4,4,4 (ablation) | ComPtr decoder · RGBT-CC | GAME0: 10.52 -> 11.01 | No clear effect |
| introduce DAB in 4 stages instead of 1 stage (ablation) | baseline · LEVIR-CD | IoU: 82.28 -> 85.41 | Improved |
| introduce CEB in 4 stages instead of 1 stage (ablation) | baseline · LEVIR-CD | IoU: 81.99 -> 84.30 | Improved |
Closest papers by meaning
- Vision Transformers for Dense Prediction (ICCV 2021 · similarity 0.88)
- Vision Transformers: From Semantic Segmentation to Dense Prediction (IJCV 2024 · similarity 0.87)
- ViT-CoMer: Vision Transformer with Convolutional Multi-scale Feature Interaction for Dense Predictions (CVPR 2024 · similarity 0.86)
- Efficient Multitask Dense Predictor via Binarization (CVPR 2024 · similarity 0.85)
- Memory Efficient Transformer Adapter for Dense Predictions (ICLR 2025 · similarity 0.84)
- Win-Win: Training High-Resolution Vision Transformers from Two Windows (ICLR 2024 · similarity 0.83)
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.
