Research map / Detection and segmentation / CNN features
Sequential Optimization for Efficient High-Quality Object Proposal Generation
Ziming Zhang, Yun Liu, Xi Chen, Yanjun Zhu, Ming‐Ming Cheng, Venkatesh Saligrama, Philip H. S. Torr
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 edge-based refinement before segmentation instead of segmentation first (ablation) | BING++ · VOC2007 test and COCO val | DR and MABO: 0.5% and 2.1% on VOC2007, 2.3% and 2.9% on COCO | Improved |
| add edge-based refinement to BING | BING · VOC2007 test | MABO: 65.4 -> 72.5 | Improved |
| add segmentation-based refinement to BING | BING · VOC2007 test | MABO: 65.4 -> 76.8 | Improved |
| add edge refinement before segmentation refinement (ablation) | S-BING · VOC2007 test | MABO: 76.8 -> 77.5 | Improved |
| resize image in segmentation refinement (ablation) | BING++ · VOC2007 test | MABO and time: 78.0 -> 78.1 MABO, 4.2 -> 3.6 ms | Improved |
| resize image in edge refinement (ablation) | BING++ · VOC2007 test | MABO and time: 78.0 -> 77.3 MABO, 4.2 -> 5.8 ms | No clear effect |
| resize in both refinements (full BING++) (ablation) | BING++ · VOC2007 test | MABO and time: 78.0 -> 77.5 MABO, 4.2 -> 2.9 ms | Mixed |
| resize in both refinements (full BING++) (ablation) | BING++ · MS COCO val | MABO and time: 57.9 -> 56.0 MABO, 11.2 -> 3.7 ms | Mixed |
| add BING++ refinement to BING | BING · VOC2007 test (IoU 0.7, 1000 proposals) | detection recall: 29.1 -> 77.3 | Improved |
| add BING++ refinement to BING | BING · VOC2007 and COCO | MABO: 18.5% and 16.7% | Improved |
| use BING++ instead of Selective Search | proposal generator · VOC2007 test | MABO and time: 81.4 -> 77.5 MABO, 10.64 -> 0.02 s | Mixed |
| vary max number of quantized scales/aspect ratios in BING (36, 121, 196) (ablation) | E-BING vs S-BING and BING++ · VOC2007 test | DR | Mixed |
Closest papers by meaning
- What Makes for Effective Detection Proposals? (TPAMI 2016 · similarity 0.87)
- Object Proposal Generation Using Two-Stage Cascade SVMs (TPAMI 2016 · similarity 0.87)
- Should All Proposals Be Treated Equally in Object Detection? (ECCV 2022 · similarity 0.86)
- Toward Scale-Invariance and Position-Sensitive Region Proposal Networks (ECCV 2018 · similarity 0.85)
- HyperNet: Towards Accurate Region Proposal Generation and Joint Object Detection (CVPR 2016 · similarity 0.85)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 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.
