Research map / Detection and segmentation / Weakly supervised detection and segmentation
Mixed Supervised Object Detection with Robust Objectness Transfer
Yan Li, Junge Zhang, Kaiqi Huang, Jianguo Zhang
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 domain-invariant (gradient-reversal) training to the objectness model (ablation) | OOM-MSD-AlexNet · ILSVRC2013 val2 (last 100 categories) | mAP (%): 18.54 -> 22.28 | Improved |
| add objectness transfer from strong categories to the weak-category detector | B-WSD-AlexNet · ILSVRC2013 val2 (last 100 categories) | mAP (%): 13.78 -> 18.54 | Improved |
| use deeper VGG16 detection network instead of AlexNet | Ours-MSD · ILSVRC2013 val2 (last 100 categories) | mAP (%): 22.28 -> 25.26 | Improved |
| use deeper VGG16 network instead of AlexNet | B-WSD · ILSVRC2013 val2 (last 100 categories) | mAP (%): 13.78 -> 11.82 | Hurt |
| fine-tune supervised strong-category detector into the WSD model (B-MSD) | B-WSD-VGG16 · ILSVRC2013 val2 (last 100 categories) | mAP (%): 11.82 -> 16.44 | Improved |
| replace fine-tuning transfer with objectness transfer | B-MSD-VGG16 · ILSVRC2013 val2 (last 100 categories) | mAP (%): 16.44 -> 25.26 | Improved |
| add domain-invariant objectness (only 10 strong categories) (ablation) | OOM-MSD-AlexNet · PASCAL VOC 2007 test (10 weak categories) | mAP (%): 14.00 -> 32.21 | Improved |
| add plain objectness transfer with only 10 strong categories | B-WSD-AlexNet · PASCAL VOC 2007 test (10 weak categories) | mAP (%): 22.63 -> 14.00 | Hurt |
| add domain-invariant objectness model instead of original objectness (ablation) | OOM-MSD-AlexNet · PASCAL VOC 2007 test (cross-dataset, 20 categories) | mAP (%): 37.65 -> 41.77 | Improved |
| add objectness transfer from 180 ILSVRC strong categories | B-WSD-AlexNet · PASCAL VOC 2007 test (cross-dataset, 20 categories) | mAP (%): 23.87 -> 41.77 | Improved |
| ensemble AlexNet and VGG16 detectors by summing scores | Ours-MSD-VGG16 · PASCAL VOC 2007 test (cross-dataset, 20 categories) | mAP (%): 47.50 -> 49.82 | Improved |
| train supervised Fast RCNN on pseudo ground truths from the ensemble | Ours-MSD-Ens · PASCAL VOC 2007 test (cross-dataset, 20 categories) | mAP (%): 49.82 -> 51.08 | Improved |
Closest papers by meaning
- Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity (NeurIPS 2021 · similarity 0.91)
- CaT: Weakly Supervised Object Detection With Category Transfer (ICCV 2021 · similarity 0.90)
- Weak-Shot Object Detection Through Mutual Knowledge Transfer (CVPR 2023 · similarity 0.89)
- Revisiting Knowledge Transfer for Training Object Class Detectors (CVPR 2018 · similarity 0.89)
- Boosting Weakly Supervised Object Detection with Progressive Knowledge Transfer (ECCV 2020 · similarity 0.88)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 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.
