Research map / Detection and segmentation / Weakly supervised detection and segmentation
Self Paced Deep Learning for Weakly Supervised Object Detection
Enver Sangineto, Moin Nabi, Dubravko R. Culibrk, Nicu Sebe
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 self-paced phase (SP) after Init | Fast-RCNN detector (h^I init) · Pascal VOC 2007 test | mAP (%): 24.74 -> 32.1 | Improved |
| add self-paced phase (SP) after Init | Fast-RCNN detector (h^P init) · Pascal VOC 2007 test | mAP (%): 31.93 -> 38.11 | Improved |
| use self-paced image selection instead of MIL on all images (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 30.8 -> 38.1 | Improved |
| use self-paced image selection instead of MIL on all images (ablation) | SP training · ILSVRC 2013 val2 | mAP (%): 8.59 -> 12.13 | Improved |
| use dynamic score selection from current network instead of static Curriculum (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 31.3 -> 38.1 | Improved |
| add class selection (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 36.9 -> 38.1 | Improved |
| add class selection (ablation) | SP training · ILSVRC 2013 val2 | mAP (%): 11.97 -> 12.13 | Improved |
| use score-based class selection instead of random class selection (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 32.7 -> 38.1 | Improved |
| use regression layer to choose latent boxes in training (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 33.5 -> 38.1 | Improved |
| use regression layer in training and test (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 30.7 -> 38.1 | Improved |
| use regression layer in training and test (ablation) | SP training · ILSVRC 2013 val2 | mAP (%): 7.57 -> 12.13 | Improved |
| restrict to one box per image via inter-classifier competition instead of one per label (SIML) (ablation) | SP training · Pascal VOC 2007 test | mAP (%): 33.80 -> 38.11 | Improved |
Closest papers by meaning
- W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection (CVPR 2018 · similarity 0.86)
- Towards Precise End-to-End Weakly Supervised Object Detection Network (ICCV 2019 · similarity 0.86)
- PreDet: Large-Scale Weakly Supervised Pre-Training for Detection (ICCV 2021 · similarity 0.86)
- Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection (CVPR 2020 · similarity 0.86)
- Active Learning Strategies for Weakly-Supervised Object Detection (ECCV 2022 · similarity 0.85)
- Multiple Instance Detection Network With Online Instance Classifier Refinement (CVPR 2017 · similarity 0.85)
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.
