Research map / Detection and segmentation / Open-set and few-shot detection
COCO-O: A Benchmark for Object Detectors under Natural Distribution Shifts
Xiaofeng Mao, Yuefeng Chen, Yao Zhu, Da Chen, Hang Su, Rong Zhang, Hui Xue
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 ResNet-50 backbone with PVTv2-B2 (ablation) | Faster R-CNN · COCO-O | COCO-O mAP: 16.4 -> 24.9 | Improved |
| use advanced neck (PAFPN, NAS-FPN) instead of FPN (ablation) | Faster R-CNN · COCO-O | Effective Robustness: -0.41 -> -0.13 (PAFPN); -2.00 (NAS-FPN) | No clear effect |
| use advanced detection head (ablation) | Faster R-CNN · COCO-O | Effective Robustness: -0.41 -> +0.62 (2_Heads) | No clear effect |
| use end-to-end transformer detectors (DETR) instead of conventional detectors | object detectors · COCO-O | Effective Robustness: -1.82 (DETR), -1.49 (Deformable DETR) | Hurt |
| add MixUp augmentation (ablation) | YOLOX-S · COCO-O | COCO-O mAP: 17.7 -> 19.8 | Improved |
| add ColorJitter augmentation (ablation) | YOLOX-S · COCO-O | COCO-O mAP: +0.3 mAP, +0.06 ER | Improved |
| pre-train backbone with SwAV instead of supervised (ablation) | Faster R-CNN · COCO-O | Effective Robustness: -0.41 -> -1.9 | Hurt |
| pre-train backbone on ImageNet-21K instead of IN-1K (ablation) | Faster R-CNN · COCO-O | Effective Robustness: -0.41 -> +1.20 | Improved |
| pre-train detector on Object365 (ablation) | Faster R-CNN · COCO-O | Effective Robustness: +0.32 | Improved |
| train for longer epochs (ablation) | Faster R-CNN, RetinaNet · COCO-O | Effective Robustness | No clear effect |
| add multi-scale training (ablation) | Faster R-CNN, RetinaNet · COCO-O | Effective Robustness: +0.26 (Faster R-CNN), +0.43 (RetinaNet) | Improved |
| add Stylized-Aug | Faster R-CNN · COCO-O | COCO-O mAP: COCO-O 16.4 -> 20.4; COCO 37.4 -> 36.1 | Mixed |
Closest papers by meaning
- OOD-CV: A Benchmark for Robustness to Out-of-Distribution Shifts of Individual Nuisances in Natural Images (ECCV 2022 · similarity 0.91)
- OOD-CV-v2 : An Extended Benchmark for Robustness to Out-of-Distribution Shifts of Individual Nuisances in Natural Images (TPAMI 2024 · similarity 0.91)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.89)
- Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion (ICLR 2025 · similarity 0.87)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.87)
- Unexplored Faces of Robustness and Out-of-Distribution: Covariate Shifts in Environment and Sensor Domains (CVPR 2024 · similarity 0.87)
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.
