Research map / Detection and segmentation / Weakly supervised detection and segmentation
End-to-End Weakly Supervised Object Detection with Sparse Proposal Evolution
Mingxiang Liao, Fang Wan, Yuan Yao, Zhenjun Han, Jialing Zou, Yuze Wang, Bailan Feng, Peng Yuan, Qixiang Ye
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 one SPR branch to seed proposals (ablation) | SPG · VOC 2007 test | mAP: 29.7 -> 42.6 | Improved |
| add one SPR branch (ablation) | SPG · VOC 2007 test | CorLoc: 57.8 -> 61.3 | Improved |
| add second and third SPR branches (ablation) | SPE · VOC 2007 test | mAP: 42.6 -> 42.9 | No clear effect |
| add seed proposal box jitter augmentation (delta_aug 0.1) (ablation) | SPE · VOC 2007 test | mAP: 42.6 -> 45.6 | Improved |
| add box jitter augmentation (delta_aug 0.1) (ablation) | SPE · VOC 2007 test | CorLoc: 61.3 -> 64.0 | Improved |
| increase jitter delta_aug from 0.1 to 0.2 (ablation) | SPE · VOC 2007 test | mAP: 45.6 -> 43.4 | Hurt |
| raise delta_seed from 0.2 to 0.4 (ablation) | SPG · VOC 2007 test | mAP: 29.7 -> 8.9 | Hurt |
| lower delta_multi from 1 to 0.5 (ablation) | SPE · VOC 2007 test | mAP: 37.8 -> 42.6 | Improved |
| share all 36 backbone blocks between SPG and SPR instead of 24 (ablation) | SPE · VOC 2007 test | mAP: 45.6 -> 32.8 | Hurt |
| share 12 blocks instead of 24 (ablation) | SPE · VOC 2007 test | mAP: 45.6 -> 43.1 | Hurt |
| use CaiT backbone with Conditional DETR instead of Faster R-CNN (fully supervised) | detector · VOC 2007 test | mAP: 81.4 -> 77.5 | Hurt |
| increase image scale from 384 to 512 | SPE · COCO 2017 | AP50: 16.3 -> 18.2 | 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.
- UWSOD — UWSOD: Toward Fully-Supervised-Level Capacity Weakly Supervised Object Detection (NeurIPS 2020) · better on 2 of 2 shared results
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
- UWSOD: Toward Fully-Supervised-Level Capacity Weakly Supervised Object Detection (NeurIPS 2020 · similarity 0.89)
- You Reap What You Sow: Using Videos to Generate High Precision Object Proposals for Weakly-Supervised Object Detection (CVPR 2019 · similarity 0.89)
- Weakly Supervised Region Proposal Network and Object Detection (ECCV 2018 · similarity 0.88)
- W2N: Switching from Weak Supervision to Noisy Supervision for Object Detection (ECCV 2022 · similarity 0.88)
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · similarity 0.88)
- SPWOOD: Sparse Partial Weakly-Supervised Oriented Object Detection (ICLR 2026 · 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.
