Research map / Detection and segmentation / CNN features
Object-Level Proposals
Jianxiang Ma, Anlong Ming, Zilong Huang, Xinggang Wang, Yu Zhou
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 occlusion-based objectness with primary occlusion edges (replacing basic edges) (ablation) | Edge Boxes pipeline · PASCAL VOC 2007 small objects, 1000 proposals, IoU 0.7 | recall: 0.142 -> 0.191 | Improved |
| add occlusion-dominated fusion of basic edges into the occlusion edge map (ablation) | Occlusion-based objectness (I) · PASCAL VOC 2007 small objects, 1000 proposals, IoU 0.7 | recall: 0.191 -> 0.204 | Improved |
| add normalization criterion on top of fusion (ablation) | Variant (II) · PASCAL VOC 2007 small objects, 1000 proposals, IoU 0.7 | recall: 0.204 -> 0.212 | Improved |
| add occlusion-based objectness with primary occlusion edges (replacing basic edges) (ablation) | Edge Boxes pipeline · PASCAL VOC 2007 large aspect ratio objects, 1000 proposals, IoU 0.7 | recall: 0.381 -> 0.453 | Improved |
| add occlusion-dominated fusion (ablation) | Occlusion-based objectness (I) · PASCAL VOC 2007 large aspect ratio objects, 1000 proposals, IoU 0.7 | recall: 0.453 -> 0.477 | Improved |
| add normalization criterion on top of fusion (ablation) | Variant (II) · PASCAL VOC 2007 large aspect ratio objects, 1000 proposals, IoU 0.7 | recall: 0.477 -> 0.483 | Improved |
| replace Edge Boxes proposals with the proposed occlusion-based proposals | Fast R-CNN VGG-M · PASCAL VOC 2007 test | mAP: 60.4 -> 61.5 | Improved |
| replace Edge Boxes proposals with the proposed ones on the boat class | Fast R-CNN VGG-M · PASCAL VOC 2007 test, boat AP | AP: 46.2 -> 45.7 | Hurt |
| use occlusion edges instead of basic edges, at a cost in runtime | Edge Boxes pipeline · PASCAL VOC 2007 runtime | runtime per image (s): 0.55s -> 0.7s | Hurt |
Closest papers by meaning
- Closing the Loop for Edge Detection and Object Proposals (AAAI 2017 · similarity 0.89)
- Object-Proposal Evaluation Protocol is 'Gameable' (CVPR 2016 · similarity 0.86)
- What Makes for Effective Detection Proposals? (TPAMI 2016 · similarity 0.86)
- Occlusion-Shared and Feature-Separated Network for Occlusion Relationship Reasoning (ICCV 2019 · similarity 0.85)
- Occlusion Boundary Detection via Deep Exploration of Context (CVPR 2016 · similarity 0.85)
- Towards Flexible 3D Perception: Object-Centric Occupancy Completion Augments 3D Object Detection (NeurIPS 2024 · similarity 0.83)
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.
