Research map / Detection and segmentation / Open-set and few-shot detection
Rectify the Regression Bias in Long-Tailed Object Detection
Ke Zhu, Minghao Fu, Jie Shao, Tianyu Liu, Jianxin Wu
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 class-agnostic regression branch (CAB) to RFS | RFS, Mask-RCNN R50-FPN · LVIS1.0 val | AP^b_r: 13.4 -> 18.0 | Improved |
| add CAB to EQLv2 | EQLv2 · LVIS1.0 val | AP^b_r: 16.1 -> 20.4 | Improved |
| add CAB to ECM | ECM · LVIS1.0 val | AP^b: 27.7 -> 29.1 | Improved |
| add CAB to ECM (segmentation rare AP) | ECM · LVIS1.0 val | AP_r (mask): 19.6 -> 19.1 | Hurt |
| use class-agnostic instead of class-specific regression head with GT labels (ablation) | CE baseline, Mask-RCNN R50-FPN · LVIS1.0 val (GT class) | AP^b_r: 34.1 -> 54.6 | Improved |
| use class-agnostic instead of class-specific head with GT labels (ablation) | CE baseline · LVIS1.0 val (GT class) | AP^b_f: 40.7 -> 40.0 | Hurt |
| mix class-agnostic and class-specific heads at alpha=0.5 instead of 0 or 1 (ablation) | RFS · LVIS1.0 val | AP^b: 24.7 -> 27.0 | Improved |
| cluster regression heads by class mean scale (k=100) (ablation) | RFS · LVIS1.0 val | AP^b: 24.7 -> 26.9 | Improved |
| cluster heads by scale with k=10 (ablation) | RFS · LVIS1.0 val | AP^b_r: 13.4 -> 12.6 | Hurt |
| merge common-class regression heads (ablation) | RFS · LVIS1.0 val | AP_r: 14.2 -> 17.7 | Improved |
| add CAB on COCO-LT at rate 3e-3 | RFS Faster-RCNN · COCO-LT | AP^b_1: 2.3 -> 6.2 | Improved |
| add CAB on balanced COCO2017 | Faster RCNN R50 · COCO2017 | AP: 37.3 -> 38.3 | Improved |
Closest papers by meaning
- Boosting Long-tailed Object Detection via Step-wise Learning on Smooth-tail Data (ICCV 2023 · similarity 0.90)
- Adaptive Class Suppression Loss for Long-Tail Object Detection (CVPR 2021 · similarity 0.89)
- Equalized Focal Loss for Dense Long-Tailed Object Detection (CVPR 2022 · similarity 0.89)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.87)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.87)
- Pursuing Better Decision Boundaries for Long-Tailed Object Detection via Category Information Amount (ICLR 2025 · 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.
