Research map / Detection and segmentation / CNN features
Beyond Quadratic: Linear-Time Change Detection with RWKV
Zhenyu Yang, Gensheng Pei, Tao Chen, Xia Yuan, Haofeng Zhang, Xiangbo Shu, Yazhou Yao
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 hierarchical RWKV encoder instead of flat VisionRWKV (ablation) | ChangeRWKV · LEVIR-CD | IoU: 81.86 -> 85.46 | Improved |
| add SE-based lightweight channel mixing (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 84.43 -> 85.46 | Improved |
| use SE module instead of heavier RWKV-style channel mixing (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 85.38 -> 85.46 IoU; 71.66G -> 33.56G FLOPs | No clear effect |
| add Spatial Fusion Module (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 82.64 -> 85.46 | Improved |
| use SFM instead of FPN-style fusion (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 83.56 -> 85.46 | Improved |
| use Cross CBAM TFM instead of SiamDiff (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 82.68 -> 85.46 | Improved |
| use Cross CBAM TFM instead of FHD (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 85.01 -> 85.46 | Improved |
| use Cross CBAM TFM instead of STRM (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 83.13 -> 85.46 | Improved |
| add Dice loss (lambda 0 -> 1.0) (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 84.95 -> 85.46 | Improved |
| increase Dice weight lambda from 1.0 to 2.0 (ablation) | ChangeRWKV-B · LEVIR-CD | IoU: 85.46 -> 85.12 | Hurt |
| scale model from Tiny to Base | ChangeRWKV · LEVIR-CD | IoU: 84.92 -> 85.46 | Improved |
| use RWKV linear-time backbone instead of Transformer | ChangeRWKV-B · LEVIR-CD, 512x512 input | FLOPs (G): 852.44G -> 134.25G | Improved |
Closest papers by meaning
- Beyond Attention: Convolutional Global Context for Remote Sensing Change Detection (ECCV 2026 · similarity 0.90)
- Changen2: Multi-Temporal Remote Sensing Generative Change Foundation Model (TPAMI 2024 · similarity 0.90)
- FM-ChangeNet: Learning Change through Pathwise Feature Transport (NeurIPS 2026 · similarity 0.88)
- The Change You Want To Detect: Semantic Change Detection In Earth Observation With Hybrid Data Generationf (CVPR 2025 · similarity 0.87)
- Feature Spectrum Learning for Remote Sensing Change Detection (CVPR 2025 · similarity 0.86)
- RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-𝑁 Ranking (NeurIPS 2026 · 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.
