Research map / Detection and segmentation / Weakly supervised detection and segmentation
TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection
Yunchao Wei, Zhiqiang Shen, Bowen Cheng, Honghui Shi, Jinjun Xiong, Jiashi Feng, Thomas Huang
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 surrounding-context completeness term (P_I - P_S) to purity-only ranking (ablation) | TS2C on OICR · VOC 2007 test | mAP: 42.2 -> 44.3 | Improved |
| add TS2C candidate pool selection to OICR | OICR · VOC 2007 test | mAP: 41.2 -> 44.3 | Improved |
| add TS2C candidate pool selection to OICR | OICR · VOC 2012 test | mAP: 37.9 -> 40.0 | Improved |
| retrain a Fast R-CNN detector on TS2C top proposals | TS2C · VOC 2007 test | mAP: 44.3 -> 48.0 | Improved |
| retrain a Fast R-CNN detector on TS2C top proposals | TS2C · VOC 2012 test | mAP: 40.0 -> 44.4 | Improved |
| use enlarge ratio 1.2 rather than 1.4 for surrounding context (ablation) | TS2C · VOC 2007 test | mAP: 42.6 -> 44.3 | Improved |
| use enlarge ratio 1.2 rather than 1.1 (ablation) | TS2C · VOC 2007 test | mAP: 43.8 -> 44.3 | Improved |
| any enlarge ratio from 1.1 to 1.4 versus OICR baseline (ablation) | TS2C · VOC 2007 test | mAP: 41.2 -> 42.6 | Improved |
| conditional average over top 50% surrounding pixels instead of top 30% (ablation) | TS2C · VOC 2007 test | mAP: 43.5 -> 44.3 | Improved |
| conditional average over top 50% surrounding pixels instead of top 70% (ablation) | TS2C · VOC 2007 test | mAP: 43.9 -> 44.3 | Improved |
| conditional average (top 50%) instead of averaging all surrounding pixels (ablation) | TS2C · VOC 2007 test | mAP: 43.8 -> 44.3 | Improved |
Later papers that compared with this method
4 later accepted papers report TS2C as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Object-Aware Instance Labeling for Weakly Supervised Object Detection (ICCV 2019 · better on 4 of 4 shared results)
- Weakly Supervised Object Detection With Segmentation Collaboration (ICCV 2019 · better on 3 of 3 shared results)
- Cap2Det: Learning to Amplify Weak Caption Supervision for Object Detection (ICCV 2019 · better on 2 of 2 shared results)
- Towards Precise End-to-End Weakly Supervised Object Detection Network (ICCV 2019 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Weakly Supervised Object Boundaries (CVPR 2016 · similarity 0.88)
- BBAM: Bounding Box Attribution Map for Weakly Supervised Semantic and Instance Segmentation (CVPR 2021 · similarity 0.87)
- Weakly Supervised Instance Segmentation using the Bounding Box Tightness Prior (NeurIPS 2019 · similarity 0.87)
- BoxTeacher: Exploring High-Quality Pseudo Labels for Weakly Supervised Instance Segmentation (CVPR 2023 · similarity 0.87)
- P2Object: Single Point Supervised Object Detection and Instance Segmentation (IJCV 2025 · similarity 0.87)
- Box2Seg: Attention Weighted Loss and Discriminative Feature Learning for Weakly Supervised Segmentation (ECCV 2020 · 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.
