Research map / Detection and segmentation / CNN features
Deformable Part Region Learning for Object Detection
Read the paperCheck 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 |
|---|---|---|---|
| Add deformable part boxes to the detector (ablation) | D-PRD with ResNet50-FPN · COCO17 val | box AP: 40.98 -> 42.20 | Helped |
| Add part classification, mask, and MaskIoU losses (ablation) | D-PRD with ResNet50-FPN · COCO17 val | box AP: 42.20 -> 42.95 | Helped |
| Add cascade refinement (ablation) | D-PRD with ResNet50-FPN · COCO17 val | box AP: 42.95 -> 44.33 | Helped |
| Use fixed rather than deformable part boxes (ablation) | Cascade D-PRD with ResNet50-FPN · COCO17 val | box AP: 45.92 -> 44.61 | Hurt |
| Use deformable parts with RDA fusion instead of fixed parts with RDA (ablation) | D-PRD · COCO17 val | box AP: 40.8 -> 42.9 | Helped |
| Use RDA fusion instead of Concat fusion with deformable parts (ablation) | D-PRD · COCO17 val | box AP: 41.7 -> 42.9 | Helped |
| Increase the part count and RDA stages from 2 parts and 2 stages to 4 parts and 3 stages (ablation) | D-PRD with ResNet50-FPN · COCO17 val | box AP: 40.78 -> 42.95 | Helped |
| Increase part count and RDA stages from 4 parts and 3 stages to 8 parts and 4 stages (ablation) | D-PRD with ResNet50-FPN · COCO17 val | box AP: 42.95 -> 42.96 | No clear effect |
| Increase part count and RDA stages from 8 parts and 4 stages to 16 parts and 5 stages (ablation) | D-PRD with ResNet50-FPN · COCO17 val | speed: 8.18 -> 6.65 | Hurt |
| Deform part boxes rather than keep them fixed (ablation) | Cascade D-PRD · COCO17 val | box AP: The paper reports an average gain of 1.2 points | Helped |
Closest papers by meaning
- Deformable ConvNets V2: More Deformable, Better Results (CVPR 2019 · similarity 0.88)
- Object Detection Based on Region Decomposition and Assembly (AAAI 2019 · similarity 0.87)
- Deformable DETR: Deformable Transformers for End-to-End Object Detection (ICLR 2021 · similarity 0.87)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018 · similarity 0.85)
- Unsupervised Part Segmentation Through Disentangling Appearance and Shape (CVPR 2021 · similarity 0.85)
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.