Research map / Detection and segmentation / Weakly supervised detection and segmentation
WSOD2: Learning Bottom-Up and Top-Down Objectness Distillation for Weakly-Supervised Object Detection
Zhaoyang Zeng, Bei Liu, Jianlong Fu, Hongyang Chao, Lei Zhang
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 bounding box regressor (ablation) | WSOD2 · PASCAL VOC 2007 test (component ablation) | mAP: 43.3 -> 45.1 | Improved |
| add NMS for seed box selection (ablation) | WSOD2 · PASCAL VOC 2007 test (component ablation) | mAP: 45.1 -> 45.9 | Improved |
| add bottom-up evidence (superpixels straddling) (ablation) | WSOD2 · PASCAL VOC 2007 test (component ablation) | mAP: 45.9 -> 48.1 | Improved |
| add alpha decay (objectness distillation schedule) (ablation) | WSOD2 · PASCAL VOC 2007 test (component ablation) | mAP: 48.1 -> 50.3 | Improved |
| add color contrast evidence (ablation) | WSOD2 · PASCAL VOC 2007 test (evidence ablation, alpha=1) | mAP: 45.9 -> 47.2 | Improved |
| add edge density evidence (ablation) | WSOD2 · PASCAL VOC 2007 test (evidence ablation, alpha=1) | mAP: 45.9 -> 47.0 | Improved |
| add multi-scale saliency evidence (ablation) | WSOD2 · PASCAL VOC 2007 test (evidence ablation, alpha=1) | mAP: 45.9 -> 47.5 | Improved |
| add superpixels straddling evidence (ablation) | WSOD2 · PASCAL VOC 2007 test (evidence ablation, alpha=1) | mAP: 45.9 -> 48.1 | Improved |
| use average of four evidences instead of superpixels straddling alone (ablation) | WSOD2 · PASCAL VOC 2007 test (evidence ablation, alpha=1) | mAP: 48.1 -> 47.9 | Hurt |
| add WSOD2 over best compared end-to-end model on VOC07 (ablation) | WSOD2 · PASCAL VOC 2007 test | mAP: 48.3 -> 53.6 | Improved |
| train on 07+12 trainval instead of 07 trainval (ablation) | WSOD2 · PASCAL VOC 2007 test | mAP: 53.6 -> 56.0 | Improved |
| use linear decay for alpha instead of other weight decay functions (ablation) | WSOD2 · PASCAL VOC 2007 test | mAP: alpha=-(n/N)+1 performs best | Improved |
Closest papers by meaning
- UWSOD: Toward Fully-Supervised-Level Capacity Weakly Supervised Object Detection (NeurIPS 2020 · similarity 0.92)
- W2N: Switching from Weak Supervision to Noisy Supervision for Object Detection (ECCV 2022 · similarity 0.91)
- Weakly Supervised Region Proposal Network and Object Detection (ECCV 2018 · similarity 0.91)
- GradingNet: Towards Providing Reliable Supervisions for Weakly Supervised Object Detection by Grading the Box Candidates (AAAI 2021 · similarity 0.90)
- Weakly Supervised Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.88)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.88)
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.
