Research map / Detection and segmentation / Open-set and few-shot detection
Exploring Generalizable Remote Sensing Change Detection via Low-Rank Exchange Adaptation of Vision Foundation Model
Mingwei Zhang, Jingtao Hu, Qiang Li, Qi Wang
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 |
|---|---|---|---|
| replace LoRA with LREA (concatenation, VCO) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 50.27 -> 66.82 | Improved |
| replace LoRA with LREA (abs. difference, VCO) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 60.21 -> 66.12 | Improved |
| replace LoRA with LREA (abs. difference, FPN) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 64.56 -> 67.05 | Improved |
| add TGFR on top of LREA (VCO) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 66.12 -> 66.90 | Improved |
| add TGFR on top of LREA (FPN) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 67.05 -> 67.96 | Improved |
| use absolute difference instead of concatenation with LoRA (VCO) (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 50.27 -> 60.21 | Improved |
| add Adapter/VPT/LoRA fine-tuning vs frozen with abs. difference (ablation) | GenCD DINOv2-L · JL2SV (cross-dataset) | mIoU: 58.00 -> 58.25 (Adapter, VCO); 61.73 -> 59.21 (Adapter, FPN) | Mixed |
| use LREA instead of LoRA on other VFMs (SAM, DINOv2-B, Panopticon, CLIP, RemoteCLIP) (ablation) | GenCD FPN · JL2SV (cross-dataset) | mIoU: 58.34 -> 60.20 (SAM-L); 53.74 -> 54.50 (CLIP-L) | Improved |
| raise LREA rank from 8 to 16 (ablation) | GenCD · JL2SV (cross-dataset) | mIoU: 66.90 -> 67.69 (VCO); 67.96 -> 68.41 (FPN) | Improved |
| raise LREA rank from 16 to 32 (ablation) | GenCD VCO · JL2SV (cross-dataset) | mIoU: 67.69 -> 59.69 | Hurt |
| insert LREA in q,k,v instead of only v (ablation) | GenCD · JL2SV (cross-dataset) | mIoU: 64.85 -> 66.90 (VCO) | Improved |
| use more TGFR tokens (10, 20) instead of one (ablation) | GenCD · JL2SV (cross-dataset) | mIoU: 66.90 -> 64.99 (10), 63.99 (20) with VCO | Hurt |
Closest papers by meaning
- Causal Decoupling Domain Generalization for Remote Sensing Change Detection (AAAI 2026 · similarity 0.91)
- Towards Generalizable Scene Change Detection (CVPR 2025 · similarity 0.88)
- Feature Spectrum Learning for Remote Sensing Change Detection (CVPR 2025 · similarity 0.88)
- The Change You Want To Detect: Semantic Change Detection In Earth Observation With Hybrid Data Generationf (CVPR 2025 · similarity 0.87)
- RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-𝑁 Ranking (NeurIPS 2026 · similarity 0.86)
- UniChange: Unifying Change Detection with Multimodal Large Language Model (CVPR 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.
