Research map / Detection and segmentation / Weakly supervised detection and segmentation
Zigzag Learning for Weakly Supervised Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use mEAS rather than region scores to rank image difficulty (ablation) | ZLDN-S · PASCAL VOC 2007 test | mAP: 34.1% -> 37.7% | Helped |
| add weighted loss for mined-instance confidence (ablation) | ZLDN-S with mEAS · PASCAL VOC 2007 test | mAP: 37.7% -> 39.1% | Helped |
| add random feature masking (ablation) | ZLDN-S with mEAS and weighted loss · PASCAL VOC 2007 test | mAP: 39.1% -> 40.7% | Helped |
| progressively add images in easy-to-difficult folds instead of training on all images at once (ablation) | ZLDN-S · PASCAL VOC 2007 test | mAP: 35.9% -> 39.1% | Helped |
| increase the number of progressive folds beyond the useful range (ablation) | ZLDN-S · PASCAL VOC 2007 test | mAP | Hurt |
| mask a small portion of mined objects with ratio τ=0.1 (ablation) | ZLDN-S · PASCAL VOC 2007 test | mAP: 40.7%; 1.6% better than the baseline without using masking | Helped |
| increase the masking ratio beyond τ=0.1 (ablation) | ZLDN-S · PASCAL VOC 2007 test | mAP | Hurt |
| use ZLDN-L rather than DSD | weakly supervised detection · PASCAL VOC 2007 test | mAP: 41.7 -> 47.6 | Helped |
| use ZLDN-L rather than DSD | weakly supervised detection · PASCAL VOC 2012 test | mAP: 38.3 -> 42.9 | Helped |
| use ZLDN-L rather than OICR | weakly supervised detection · PASCAL VOC 2007 trainval | CorLoc: 60.6% -> 61.2% | Helped |
| use ZLDN-L rather than OICR | weakly supervised detection · PASCAL VOC 2012 trainval | CorLoc: 62.1% vs. 61.5% | Hurt |
Closest papers by meaning
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.87)
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · similarity 0.86)
- You Reap What You Sow: Using Videos to Generate High Precision Object Proposals for Weakly-Supervised Object Detection (CVPR 2019 · similarity 0.86)
- Track and Transfer: Watching Videos to Simulate Strong Human Supervision for Weakly-Supervised Object Detection (CVPR 2016 · similarity 0.86)
- DAP: Detection-Aware Pre-Training With Weak Supervision (CVPR 2021 · similarity 0.86)
- Multi-Evidence Filtering and Fusion for Multi-Label Classification, Object Detection and Semantic Segmentation Based on Weakly Supervised Learning (CVPR 2018 · similarity 0.86)
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.