Research map / Detection and segmentation / CNN features
Mask-Guided Attention Network for Occluded Pedestrian Detection
Yanwei Pang, Jin Xie, Muhammad Haris Khan, Rao Muhammad Anwer, Fahad Shahbaz Khan, Ling Shao
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 mask-guided attention branch (Lmask) (ablation) | SPD baseline VGG-16 · CityPersons val HO | log-average miss rate (%): 57.0 -> 52.7 | Improved |
| add mask-guided attention branch (Lmask) (ablation) | SPD baseline VGG-16 · CityPersons val R | log-average miss rate (%): 13.8 -> 11.9 | Improved |
| add occlusion-sensitive loss (Locc) alone (ablation) | SPD baseline VGG-16 · CityPersons val HO | log-average miss rate (%): 57.0 -> 55.6 | Improved |
| add Locc on top of mask-guided attention (ablation) | MGAN · CityPersons val HO | log-average miss rate (%): 52.7 -> 51.7 | Improved |
| add Locc on top of mask-guided attention (ablation) | MGAN · CityPersons val R | log-average miss rate (%): 11.9 -> 11.5 | Improved |
| use coarse box-level masks instead of dense pixel-wise masks (ablation) | MGAN · CityPersons val HO | log-average miss rate (%): 51.7 -> 52.7 | No clear effect |
| use coarse box-level masks instead of dense pixel-wise masks (ablation) | MGAN · CityPersons val R | log-average miss rate (%): 11.2 -> 11.9 | No clear effect |
| replace CA attention with mask-guided attention (ablation) | MGAN without Locc · CityPersons val HO | log-average miss rate (%): 54.5 -> 52.7 | Improved |
| replace CA-VBB attention with mask-guided attention (ablation) | MGAN without Locc · CityPersons val HO | log-average miss rate (%): 54.1 -> 52.7 | Improved |
| add MGAN to baseline for small heavily occluded pedestrians | SPD baseline · CityPersons val HO 50-75 px | log-average miss rate (%): 66.3 -> 61.7 | Improved |
| add MGAN to baseline for large heavily occluded pedestrians | SPD baseline · CityPersons val HO >125 px | log-average miss rate (%): 43.1 -> 37.6 | Improved |
| replace Bi-Box with MGAN | detector VGG-16 · CityPersons val HO | log-average miss rate (%): 44.2 -> 39.4 | Improved |
Closest papers by meaning
- Occluded Pedestrian Detection Through Guided Attention in CNNs (CVPR 2018 · similarity 0.95)
- Bi-box Regression for Pedestrian Detection and Occlusion Estimation (ECCV 2018 · similarity 0.92)
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.92)
- Occlusion-aware R-CNN: Detecting Pedestrians in a Crowd (ECCV 2018 · similarity 0.91)
- Discriminative Feature Transformation for Occluded Pedestrian Detection (ICCV 2019 · similarity 0.90)
- High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (CVPR 2019 · similarity 0.90)
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.
