Research map / Detection and segmentation / CNN features
Graininess-Aware Deep Feature Learning for Pedestrian Detection
Chunze Lin, Jiwen Lu, Gang Wang, Jie Zhou
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 |
|---|---|---|---|
| use multi-layer detection instead of conv_fc7 only (ablation) | GDFL backbone · Caltech Reasonable | miss rate: 16.86 -> 9.44 | Improved |
| use multi-layer detection instead of conv_fc7 only (ablation) | GDFL backbone · Caltech Heavy Occlusion | miss rate: 53.44 -> 50.21 | Improved |
| add single-scale pedestrian attention (ablation) | multi-layer backbone · Caltech Heavy Occlusion | miss rate: 50.21 -> 47.60 | Improved |
| use scale-aware attention instead of single-scale attention (ablation) | GDFL · Caltech Heavy Occlusion | miss rate: 47.60 -> 44.68 | Improved |
| add instance-sensitive weight in attention loss (ablation) | GDFL · Caltech Heavy Occlusion | miss rate: 47.69 -> 44.68 | Improved |
| add instance-sensitive weight in attention loss (ablation) | GDFL · Caltech Reasonable | miss rate: 9.59 -> 8.44 | Improved |
| add ZIZOM on attention-encoded conv4_3 (ablation) | GDFL · Caltech Heavy Occlusion | miss rate: 44.68 -> 43.18 | Improved |
| add ZIZOM on attention-encoded conv4_3 (ablation) | GDFL · Caltech Medium | miss rate: 33.45 -> 32.50 | Improved |
| add ZIZOM on conv5_3 in addition to conv4_3 (ablation) | GDFL · Caltech | miss rate: 43.18 -> 42.86 (Heavy), 7.36 -> 8.01 (Reasonable) | No clear effect |
| add ZIZOM on conv4_3 without attention (ablation) | multi-layer backbone · Caltech Heavy Occlusion | miss rate: 50.21 -> 45.73 | Improved |
| up-sample KITTI input by 1.5 times (384x1280 to 576x1920) (ablation) | GDFL · KITTI Hard | mAP: 60.07 -> 66.86 | Improved |
Closest papers by meaning
- Mask-Guided Attention Network for Occluded Pedestrian Detection (ICCV 2019 · similarity 0.89)
- High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (CVPR 2019 · similarity 0.88)
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.87)
- Occluded Pedestrian Detection Through Guided Attention in CNNs (CVPR 2018 · similarity 0.86)
- From Handcrafted to Deep Features for Pedestrian Detection: A Survey (TPAMI 2021 · similarity 0.86)
- Bi-box Regression for Pedestrian Detection and Occlusion Estimation (ECCV 2018 · 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.
