Research map / Detection and segmentation / Weakly supervised detection and segmentation
CaT: Weakly Supervised Object Detection With Category Transfer
Tianyue Cao, Lianyu Du, Xiaoyun Zhang, Siheng Chen, Ya Zhang, Yan-Feng Wang
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 double-supervision mean teacher (DSMT) (ablation) | weakly supervised mean teacher baseline · Pascal VOC 2007 test (COCO-65 source) | mAP / CorLoc: 35.2 -> 60.3 mAP; 55.5 -> 77.8 CorLoc | Improved |
| add semantic graph convolutional network on top of DSMT (ablation) | CaT · Pascal VOC 2007 test (COCO-65 source) | mAP / CorLoc: 60.3 -> 63.0 mAP; 77.8 -> 80.0 CorLoc | Improved |
| increase the number of overlapping categories n (ablation) | CaT · Pascal VOC 2007 test | mAP | Improved |
| remove images (not only annotations) of overlapping categories from COCO source (ablation) | CaT · Pascal VOC 2007 test | mAP / CorLoc: 58.0 -> 59.2 mAP; 73.8 -> 75.9 CorLoc | Improved |
| add 5 overlapping categories (COCO-65 instead of COCO-60) (ablation) | CaT · Pascal VOC 2007 test | mAP / CorLoc: 58.0 -> 63.5 mAP; 73.8 -> 80.3 CorLoc | Improved |
| add category transfer (CaT_0) vs best pure WSOD CASD | CaT · Pascal VOC 2007 test | mAP / CorLoc: 56.8 -> 58.0 mAP; 70.4 -> 73.8 CorLoc | Improved |
| CaT_0 vs previous transfer method Boosting (COCO-60) | CaT · Pascal VOC 2007 test | mAP / CorLoc: 55.2 -> 58.0 mAP; 72.4 -> 73.8 CorLoc | Improved |
| use similarity inter-dataset edges rather than hand-crafted edges (ablation) | CaT · KITTI test | mAP: 80.6 mAP with similarity edges | Improved |
| combine similarity and hand-crafted inter-dataset edges (ablation) | CaT · KITTI test | CorLoc: 58.9 CorLoc with both edges | Mixed |
| vary fully-supervised loss weight lambda_full away from 0.5 (ablation) | CaT · Pascal VOC 2007 test | mAP | Hurt |
| increase graph threshold tau away from 0.4 (sparser graph) (ablation) | CaT · Pascal VOC 2007 test | mAP | Hurt |
Closest papers by meaning
- Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity (NeurIPS 2021 · similarity 0.92)
- Boosting Weakly Supervised Object Detection with Progressive Knowledge Transfer (ECCV 2020 · similarity 0.91)
- Large Scale Semi-Supervised Object Detection Using Visual and Semantic Knowledge Transfer (CVPR 2016 · similarity 0.91)
- Visual and Semantic Knowledge Transfer for Large Scale Semi-Supervised Object Detection (TPAMI 2017 · similarity 0.91)
- Weakly Supervised Object Detection Using Proposal- and Semantic-Level Relationships (TPAMI 2020 · similarity 0.90)
- Relational Matching for Weakly Semi-Supervised Oriented Object Detection (CVPR 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.
