Research map / Detection and segmentation / CNN features
Large Selective Kernel Network for Remote Sensing Object Detection
Yuxuan Li, Qibin Hou, Zhaohui Zheng, Ming-Ming Cheng, Jian Yang, Xiang Li
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 |
|---|---|---|---|
| decompose a large kernel into two depth-wise kernels instead of one single kernel (ablation) | LSKNet-T · DOTA-v1.0 | mAP / FPS: 80.66 -> 80.91 mAP; 18.6 -> 20.5 FPS | Improved |
| decompose into three kernels rather than two (ablation) | LSKNet-T · DOTA-v1.0 | mAP / FPS: 80.91 -> 80.77 mAP; 20.5 -> 19.2 FPS | Hurt |
| use spatial kernel selection rather than none (ablation) | LSKNet-T · DOTA-v1.0 | mAP: 80.94 -> 81.31 | Improved |
| use channel selection (SKNet-style) instead of spatial selection (ablation) | LSKNet-T · DOTA-v1.0 | mAP: 81.31 -> 80.57 | Hurt |
| use receptive field of 23 instead of 11 or 39 (ablation) | LSKNet-T · DOTA-v1.0 | mAP: 80.80 (RF 11), 80.84 (RF 39) -> 80.94 (RF 23) | Improved |
| use both max and average pooling in spatial selection (ablation) | LSKNet-T · DOTA-v1.0 | mAP: 81.23 (max), 81.12 (avg) -> 81.31 | Improved |
| replace ResNet-18 backbone with LSKNet-T across detection frameworks | Oriented RCNN, RoI Transformer, S2A-Net, R3Det · DOTA-v1.0 | mAP: 79.27 -> 81.31 (O-RCNN); 74.16 -> 78.39 (R3Det) | Improved |
| use LSKNet-S instead of other large kernel or selective attention backbones | Oriented RCNN · DOTA-v1.0 | mAP: 81.15 (VAN-B1) -> 81.48 | Improved |
| use LSKNet-S backbone vs O-RCNN with prior best | Oriented RCNN · FAIR1M-v1.0 | mAP: 45.60 -> 47.87 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ReDet — ReDet: A Rotation-Equivariant Detector for Aerial Object Detection (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LSKNet: A Foundation Lightweight Backbone for Remote Sensing (IJCV 2024 · similarity 0.95)
- Poly Kernel Inception Network for Remote Sensing Detection (CVPR 2024 · similarity 0.87)
- PKINet-v2: Towards Powerful and Efficient Poly-Kernel Remote Sensing Object Detection (ECCV 2026 · similarity 0.87)
- Active Learning Meets Foundation Models: Fast Remote Sensing Data Annotation for Object Detection (ICCV 2025 · similarity 0.85)
- Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community (AAAI 2025 · similarity 0.84)
- LSK3DNet: Towards Effective and Efficient 3D Perception with Large Sparse Kernels (CVPR 2024 · similarity 0.84)
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.
