Research map / Detection and segmentation / Weakly supervised detection and segmentation
Large Scale Semi-Supervised Object Detection Using Visual and Semantic Knowledge Transfer
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 |
|---|---|---|---|
| Adapt category-specific classifier weights using similarity-informed transfer | LSDA · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 16.33 -> 19.02 | Helped |
| Combine visual similarity and semantic relatedness | LSDA · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 16.15 -> 20.03; ↑3.88 | Helped |
| Use visual similarity to weight transfer from all 100 source categories instead of LSDA's nearest-neighbor weighting | LSDA · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 16.33 -> 19.02 | Helped |
| Use semantic sparse representation for transfer instead of semantic weighted nearest neighbors (ablation) | Semantic transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 18.32 -> 19.04 | Helped |
| Use visual rather than semantic transfer (ablation) | Similarity-based knowledge transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 19.02 vs. 19.04 | No clear effect |
| Use visual similarity rather than LSDA's fc8-weight Euclidean distance to choose neighbors (ablation) | LSDA · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 16.33 -> 19.02 | Helped |
| Weight each source category by its visual similarity rather than average the transfer from 5 neighbors (ablation) | Visual transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 17.42 -> 17.59 | Helped |
| Weight each source category by its visual similarity rather than average the transfer from 10 neighbors (ablation) | Visual transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 17.62 -> 18.41 | Helped |
| Increase the visual neighbor set from 5 to 10 while averaging transfer weights (ablation) | Visual transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 17.42 -> 17.62 | Helped |
| Increase the visual neighbor set from 10 to 100 with weighted transfer (ablation) | Visual transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 18.41 -> 19.02 | Helped |
| Use semantic sparse reconstruction instead of semantic weighted-100 transfer (ablation) | Semantic transfer · ILSVRC2013 val2, all 200 categories | mAP: 23.28 -> 23.66 | Helped |
| Combine visual and semantic similarity rather than use either alone (ablation) | Similarity-based knowledge transfer · ILSVRC2013 val2, weakly labeled 100 categories | mAP: 19.02 -> 20.03 | Helped |
Closest papers by meaning
- Revisiting Knowledge Transfer for Training Object Class Detectors (CVPR 2018 · similarity 0.90)
- Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity (NeurIPS 2021 · similarity 0.90)
- Universal-RCNN: Universal Object Detector via Transferable Graph R-CNN (AAAI 2020 · similarity 0.89)
- Open-Vocabulary Object Detection Using Captions (CVPR 2021 · similarity 0.89)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.87)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.87)
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.