Research map / Detection and segmentation / Weakly supervised detection and segmentation
Towards Precise End-to-End Weakly Supervised Object Detection Network
Ke Yang, Dongsheng Li, Yong Dou
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 guided attention module (GAM) to the MIL detector (ablation) | MIL detector (OICR) · VOC 2007 test | mAP: 41.3 -> 43.3 | Improved |
| add GAM to the MIL detector (ablation) | MIL detector (OICR) · VOC 2007 trainval | CorLoc: 60.6 -> 61.4 | Improved |
| train regression branch jointly instead of a separate Fast R-CNN phase (ablation) | MIL detector · VOC 2007 test | mAP: 45.9 -> 47.3 | Improved |
| train regression branch jointly instead of a separate Fast R-CNN phase (ablation) | MIL detector · VOC 2007 trainval | CorLoc: 64.5 -> 65.8 | Improved |
| add GAM on top of joint regression (ablation) | MIL+REG · VOC 2007 test | mAP: 47.3 -> 48.6 | Improved |
| add GAM on top of joint regression (ablation) | MIL+REG · VOC 2007 trainval | CorLoc: 65.8 -> 66.8 | Improved |
| add joint regression branch to MIL detector (ablation) | MIL detector · VOC 2007 test | mAP: 41.3 -> 47.3 | Improved |
| add separate Fast R-CNN second phase to MIL detector (ablation) | MIL detector · VOC 2007 test | mAP: 41.3 -> 45.9 | Improved |
| use PCL instead of OICR as the MIL base with GAM+REG | MIL+GAM+REG · VOC 2007 test | mAP: 48.6 -> 51.5 | Improved |
| use PCL instead of OICR as the MIL base with GAM+REG | MIL+GAM+REG · VOC 2012 test | mAP: 46.8 -> 45.6 | Hurt |
| single joint model versus OICR ensemble with Fast R-CNN | OICR · VOC 2007 test | mAP: 47.0% mAP Versus 48.6% mAP | Improved |
| single joint model versus OICR ensemble with Fast R-CNN | OICR · VOC 2007 trainval | CorLoc: 60.6% CorLoc Versus 66.8% CorLoc | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TS2C — TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection (ECCV 2018) · 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
- Multiple Instance Detection Network With Online Instance Classifier Refinement (CVPR 2017 · similarity 0.91)
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.90)
- PCL: Proposal Cluster Learning for Weakly Supervised Object Detection (TPAMI 2018 · similarity 0.90)
- Weakly Supervised Object Localization with Multi-Fold Multiple Instance Learning (TPAMI 2016 · similarity 0.90)
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · similarity 0.90)
- Weakly Supervised Object Detection With Segmentation Collaboration (ICCV 2019 · similarity 0.89)
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.
