Research map / Detection and segmentation / CNN features
Equalized Focal Loss for Dense Long-Tailed Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the shared focal-loss modulator with category-specific focusing and weighting factors | one-stage detector with improved baseline · LVIS v1 val | AP: 25.7 to 27.5 | Helped |
| add category-specific focusing factor (ablation) | focal-loss improved baseline · LVIS v1 val | AP: 25.7 to 26.2 | Helped |
| add weighting factor (ablation) | focal-loss improved baseline · LVIS v1 val | AP: 25.7 to 26.1 | Helped |
| combine focusing and weighting factors (ablation) | focal-loss improved baseline · LVIS v1 val | AP: 25.7 to 27.5 | Helped |
| set the EFL scale factor s to 1 (ablation) | EFL · LVIS v1 val | AP: 25.7 to 26.3 | Helped |
| set the EFL scale factor s to 2 (ablation) | EFL · LVIS v1 val | AP: 25.7 to 26.6 | Helped |
| set the EFL scale factor s to 4 (ablation) | EFL · LVIS v1 val | AP: 25.7 to 27.3 | Helped |
| increase the EFL scale factor from 8 to 12 (ablation) | EFL · LVIS v1 val | AP: 27.5 to 26.5 | Hurt |
| add stabilized training settings (ablation) | EFL one-stage detector · LVIS v1 val | AP: +0.3% AP | Helped |
| replace the centerness branch with an IoU branch (ablation) | EFL one-stage detector with stabilized settings · LVIS v1 val | AP: +1.0% AP | Helped |
| increase the anchor scale (ablation) | EFL one-stage detector with stabilized settings · LVIS v1 val | AP: +0.7% AP | Helped |
| apply EFL to RetinaNet | RetinaNet · LVIS v1 val | AP: 18.5 to 20.5 | Helped |
Closest papers by meaning
- Equalization Loss v2: A New Gradient Balance Approach for Long-Tailed Object Detection (CVPR 2021 · similarity 0.92)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.88)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.88)
- Pursuing Better Decision Boundaries for Long-Tailed Object Detection via Category Information Amount (ICLR 2025 · similarity 0.88)
- Fractal Calibration for Long-tailed Object Detection (CVPR 2025 · similarity 0.87)
- OFBD: Object-Focused Background Debiasing for Long-Tailed Learning (NeurIPS 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.