Research map / Detection and segmentation / Open-set and few-shot detection
Towards Large-Scale Small Object Detection: Survey and Benchmarks
Gong Cheng, Xiang Yuan, Xiwen Yao, Kebing Yan, Qinghua Zeng, Xingxing Xie, Junwei Han
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 RFLA Gaussian receptive field label assignment (ablation) | Faster RCNN · SODA-D test | AP: 28.9 -> 29.7 AP, AP_eS 13.9 -> 13.2 | Mixed |
| replace anchor matching with FreeAnchor (ablation) | RetinaNet · SODA-D test | AP: 28.2 -> 29.6 | Improved |
| replace anchor assignment with PAA (ablation) | RetinaNet · SODA-D test | AP: 28.2 -> 29.7 AP, AP_eS 11.9 -> 12.0 | Mixed |
| use ATSS dynamic label assignment (ablation) | FCOS · SODA-D test | AP: 23.9 -> 26.8 | Improved |
| use AutoAssign re-weighted assignment (ablation) | FCOS · SODA-D test | AP: 23.9 -> 25.5 | Improved |
| add GHM loss (ablation) | RetinaNet · SODA-D test | AP: 28.2 -> 28.4, AP_eS 11.9 -> 12.5 | No clear effect |
| add GFL loss (ablation) | ATSS · SODA-D test | AP: 26.8 -> 29.0 | Improved |
| use ConvNeXt-T backbone instead of ResNet-50 (ablation) | Cascade RCNN · SODA-D test | AP: 31.2 -> 34.3 | Improved |
| use ResNet-101 instead of ResNet-50 (ablation) | Faster RCNN · SODA-D test | AP: 28.9 -> 28.7 | No clear effect |
| use Swin-T backbone instead of ResNet-50 (ablation) | FCOS · SODA-D test | AP: 23.9 -> 29.2 | Improved |
| increase proposals per patch from 2000 to 8000 (ablation) | Oriented RCNN · SODA-A test | AP: 34.4 -> 34.1 AP, 12.3 -> 8.7 FPS | No clear effect |
| increase proposals per patch from 1000 to 2000 (ablation) | Oriented RCNN · SODA-A test | AP: 33.9 -> 34.4 | Improved |
Closest papers by meaning
- Oriented Tiny Object Detection: A Dataset, Benchmark, and Dynamic Unbiased Learning (TPAMI 2025 · similarity 0.91)
- Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (TPAMI 2021 · similarity 0.91)
- Salient Object Detection in the Deep Learning Era: An In-Depth Survey (TPAMI 2021 · similarity 0.91)
- "A Simple Approach and Benchmark for 21,000-Category Object Detection" (ECCV 2022 · similarity 0.91)
- Deep Learning for Generic Object Detection: A Survey (IJCV 2019 · similarity 0.90)
- Learning Rich Features at High-Speed for Single-Shot Object Detection (ICCV 2019 · similarity 0.90)
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.
