Research map / Detection and segmentation / Weakly supervised detection and segmentation
Salvage of Supervision in Weakly Supervised Object Detection
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 pseudo groundtruth filtering (PGF) (ablation) | SoS-WSOD · VOC2007 | mAP50: 54.1 -> 57.6 | Helped |
| Add semi-supervised detection (SSOD) (ablation) | SoS-WSOD · VOC2007 | mAP50: 57.6 -> 62.7 | Helped |
| Add pseudo groundtruth filtering (PGF) (ablation) | SoS-WSOD · VOC2012 | mAP50: 51.8 -> 53.9 | Helped |
| Add semi-supervised detection (SSOD) (ablation) | SoS-WSOD · VOC2012 | mAP50: 53.9 -> 59.6 | Helped |
| Add pseudo groundtruth filtering (PGF) (ablation) | SoS-WSOD · MS-COCO | mAP50: 23.6 -> 27.5 | Helped |
| Add semi-supervised detection (SSOD) (ablation) | SoS-WSOD · MS-COCO | mAP50: 27.5 -> 30.6 | Helped |
| Use the proposed clean/noisy data split rule rather than no split rule (ablation) | SoS-WSOD stage 3 · VOC2007 | mAP50: 59.6 -> 62.7 | Helped |
| Use random splitting instead of the proposed split rule (ablation) | SoS-WSOD stage 3 · VOC2007 | mAP50: 62.7 -> 61.1 | Hurt |
| Double the WSOD-stage training iterations (ablation) | SoS-WSOD · VOC2007 | mAP50: 54.1% to 52.7% | Hurt |
| Increase the labeled subset size from K=2000 to K=3000 (ablation) | SoS-WSOD stage 3 · VOC2007 | mAP50: 62.7 -> 62.3 | Hurt |
| Use PGF rather than top-one pseudo-label mining (ablation) | SoS-WSOD pseudo-FSOD · VOC2007 | mAP50: 53.4 -> 57.6 | Helped |
| Use PGF rather than W2F pseudo-label mining (ablation) | SoS-WSOD pseudo-FSOD · MS-COCO | mAP50: 24.1 -> 27.5 | Helped |
Closest papers by meaning
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.91)
- UWSOD: Toward Fully-Supervised-Level Capacity Weakly Supervised Object Detection (NeurIPS 2020 · similarity 0.91)
- GradingNet: Towards Providing Reliable Supervisions for Weakly Supervised Object Detection by Grading the Box Candidates (AAAI 2021 · similarity 0.89)
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.89)
- Comprehensive Attention Self-Distillation for Weakly-Supervised Object Detection (NeurIPS 2020 · similarity 0.88)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · 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.