Research map / Detection and segmentation / Weakly supervised detection and segmentation
Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity
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 |
|---|---|---|---|
| add a coarse mask prior at the last backbone layer (ablation) | Plain weak-shot detection framework · VOC-20 test (COCO-60 source, single-scale) | mAP: 58.2 -> 59.7 | Helped |
| add a coarse mask prior at the penultimate backbone layer (ablation) | Plain weak-shot detection framework · VOC-20 test (COCO-60 source, single-scale) | mAP: 58.2 -> 59.1 | Helped |
| place the mask prior at the penultimate rather than the last backbone layer (ablation) | TraMaS mask-prior variant · VOC-20 test (COCO-60 source, single-scale) | mAP: 59.7 -> 59.1 | Hurt |
| add SimNet-based similarity weighting to suppress noisy pseudo boxes (ablation) | Plain weak-shot detection framework · VOC-20 test (COCO-60 source, single-scale) | mAP: 58.2 -> 59.5 | Helped |
| use cosine similarity instead of SimNet for similarity weighting (ablation) | TraMaS similarity-weighting variant · VOC-20 test (COCO-60 source, single-scale) | mAP: 59.5 -> 58.8 | Hurt |
| combine last-layer mask prior with SimNet similarity weighting (ablation) | Plain weak-shot detection framework · VOC-20 test (COCO-60 source, single-scale) | mAP: 58.2 -> 60.9 | Helped |
| increase source data from 20% to 50% of COCO-60 | TraMaS · VOC-20 test (single-scale) | mAP: 59.4% -> 59.8% | Helped |
| provide full box supervision for categories also present in the weakly labeled target data | TraMaS+distill (ResNet50) · VOC-20 test (mixed vs novel categories) | mAP: 69.9% (mixed categories) v.s. 62.4% (novel categories) | Helped |
| use TraMaS rather than the strongest WSHOD baseline | VOC-20 test (COCO-60 source) · VOC-20 test | mAP: 60.2 -> 62.1 | Helped |
| use TraMaS rather than the strongest WSHOD baseline | VOC-20 test (ILSVRC-179 source) · VOC-20 test | mAP: 56.5 -> 58.3 | Helped |
Closest papers by meaning
- Large Scale Semi-Supervised Object Detection Using Visual and Semantic Knowledge Transfer (CVPR 2016 · similarity 0.90)
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.90)
- Revisiting Knowledge Transfer for Training Object Class Detectors (CVPR 2018 · similarity 0.90)
- Weak-Shot Object Detection Through Mutual Knowledge Transfer (CVPR 2023 · similarity 0.90)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.86)
- Relational Matching for Weakly Semi-Supervised Oriented Object Detection (CVPR 2024 · 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.