Research map / Detection and segmentation / Weakly supervised detection and segmentation
DLWL: Improving Detection for Lowshot Classes With Weakly Labelled Data
Check 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 |
|---|---|---|---|
| add weakly labelled YFCC100M images and constrained proposal-label assignment | FRCNN with ResNet-50 · LVIS rare classes | AP-r: 10.84 ± 0.76 -> 14.21 ± 1.03 | Helped |
| add weakly labelled YFCC100M images and constrained proposal-label assignment | FRCNN with ResNeXt-101-32x8d · LVIS rare classes | AP-r: 12.73 ± 1.18 -> 17.36 ± 0.80 | Helped |
| add weakly labelled examples using DLWL | FRCNN · COCO lowshot highshot classes, N = 10 | AP: 35.7 -> 35.7 | No clear effect |
| add per-image class counts as weak supervision (ablation) | DLWL · COCO lowshot, N = 10 | AP: 17.1 -> 18.2 | Helped |
| set Nc to 1 rather than estimating the average instance count (ablation) | DLWL · COCO lowshot, N = 10 | AP: 17.1 -> 13.2 | Hurt |
| set Nc to 2 rather than estimating the average instance count (ablation) | DLWL · COCO lowshot, N = 10 | AP: 17.1 -> 15.8 | Hurt |
| set Nc to 3 rather than estimating the average instance count (ablation) | DLWL · COCO lowshot, N = 10 | AP: 17.1 -> 16.5 | Hurt |
| set Nc to 1 rather than the default count of 3 (ablation) | DLWL · COCO lowshot, N = 0 | AP: 12.1 -> 9.8 | Hurt |
| use omni-weak instead of DLWL | FRCNN · LVIS rare classes | AP-r: 14.21 ± 1.03 -> 12.88 ± 1.23 | Hurt |
| use DLWL rather than PCL | VGG-16 weakly supervised detector · COCO14 | AP50: 19.6 -> 19.5 | No clear effect |
| use DLWL rather than PCL | VGG-16 weakly supervised detector · COCO14 | AP: 9.2 -> 9.2 | No clear effect |
Closest papers by meaning
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.89)
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · similarity 0.88)
- Multi-Evidence Filtering and Fusion for Multi-Label Classification, Object Detection and Semantic Segmentation Based on Weakly Supervised Learning (CVPR 2018 · similarity 0.88)
- Partial Weakly-Supervised Oriented Object Detection (CVPR 2026 · similarity 0.88)
- UniT: Unified Knowledge Transfer for Any-Shot Object Detection and Segmentation (CVPR 2021 · similarity 0.88)
- Open-Vocabulary Object Detection Using Captions (CVPR 2021 · 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.