Research map / Detection and segmentation / Open-set and few-shot detection
A Shape Transformation-based Dataset Augmentation Framework for Pedestrian Detection
Zhe Chen, Wanli Ouyang, Tongliang Liu, Dacheng Tao
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 shape-guided warping loss L_shape to baseline (ablation) | FPN detector · Caltech validation split (ablation) | miss rate (%): 10.77 -> 10.58 | Improved |
| add shape constraining operation (ablation) | STDA · Caltech validation split (ablation) | miss rate (%): 10.58 -> 10.31 | Improved |
| add cyclic reconstruction loss (ablation) | STDA · Caltech validation split (ablation) | miss rate (%): 10.31 -> 9.03 | Improved |
| add adversarial loss (ablation) | STDA · Caltech validation split (ablation) | miss rate (%): 9.03 -> 8.56 | Improved |
| add environment-aware blending map (ablation) | STDA · Caltech validation split (ablation) | miss rate (%): 8.56 -> 7.73 | Improved |
| add hard positive mining loss (ablation) | STDA · Caltech validation split (ablation) | miss rate (%): 7.73 -> 7.49 | Improved |
| augment Caltech with STDA | FPN detector · Caltech test | miss rate (%): 10.4 baseline; around 30% improvement | Improved |
| add multi-scale testing to STDA | FPN w STDA · Caltech test | miss rate (%): 30% -> 38% improvement | Improved |
| augment CityPersons with STDA | FPN detector · CityPersons val | miss rate (%) Reasonable: 13.9 -> 11.0 | Improved |
| add multi-scale testing to STDA | FPN w STDA · CityPersons val | miss rate (%) Reasonable: 11.0 -> 10.2 | Improved |
| augment Caltech with STDA | FPN detector · Caltech test, heavy occlusion | miss rate (%): 59.3 -> 41.9 | Improved |
| augment Caltech with STDA | FPN detector · Caltech test, partial occlusion | miss rate (%): 22.9 -> 17.2 | Improved |
Closest papers by meaning
- Artificial Dummies for Urban Dataset Augmentation (AAAI 2021 · similarity 0.86)
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.85)
- Discriminative Feature Transformation for Occluded Pedestrian Detection (ICCV 2019 · similarity 0.85)
- Bi-box Regression for Pedestrian Detection and Occlusion Estimation (ECCV 2018 · similarity 0.84)
- Generalizable Pedestrian Detection: The Elephant in the Room (CVPR 2021 · similarity 0.84)
- High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (CVPR 2019 · similarity 0.84)
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.
