Research map / Detection and segmentation / CNN features
Progressive Refinement Network for Occluded Pedestrian Detection
Xiaolin Song Kaili Zhao Wen-Sheng Chu Honggang Zhang Jun Guo
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 visible-part estimation and anchor calibration before the full-body refiner (ablation) | PRNet-F · CityPersons val | MR-2 R: 15.6 -> 11.4 | Improved |
| add visible-part estimation and anchor calibration before the full-body refiner (ablation) | PRNet-F · CityPersons val | MR-2 HO: 45.7 -> 45.3 | Improved |
| use VE+AC alone instead of a standalone full-body detector (ablation) | PRNet-F · CityPersons val | MR-2 R / HO: R 15.6 -> 11.7, HO 45.7 -> 51.3 | Mixed |
| add full-body refinement after VE+AC (ablation) | PRNet-VA · CityPersons val | MR-2 HO: 51.3 -> 45.3 | Improved |
| add full-body refinement after VE+AC (ablation) | PRNet-VA · CityPersons val | MR-2 R: 11.7 -> 11.4 | Improved |
| use anchor calibration instead of a learned box regressor from visible to full-body boxes (ablation) | PRNet-VAF · CityPersons val | MR-2 R: 12.6 -> 11.4 | Improved |
| use anchor calibration instead of a learned box regressor (ablation) | PRNet-VAF · CityPersons val | MR-2 HO: 44.7 -> 45.3 | Hurt |
| add occlusion loss (ablation) | PRNet-VAF · CityPersons val | MR-2 R / HO: R 11.4 -> 11.0, HO 45.3 -> 45.7 | Mixed |
| add Receptive Field Backfeed (RFB) (ablation) | PRNet-VAF · CityPersons val | MR-2 R / HO: R 11.4 -> 11.6, HO 45.3 -> 44.9 | Mixed |
| add both occlusion loss and RFB (ablation) | PRNet-VAF · CityPersons val | MR-2 R / HO: R 11.4 -> 10.8, HO 45.3 -> 42.0 | Improved |
| PRNet versus ALFNet on the Heavy subset | PRNet · CityPersons val | MR-2 Heavy: 51.9 -> 53.3 | Hurt |
| PRNet versus ALFNet cross-dataset to ETH | PRNet · ETH (trained on CityPersons) | MR-2 R+HO: 31.1 -> 27.0 | Improved |
Closest papers by meaning
- Improving Occlusion and Hard Negative Handling for Single-Stage Pedestrian Detectors (CVPR 2018 · similarity 0.90)
- Mask-Guided Attention Network for Occluded Pedestrian Detection (ICCV 2019 · similarity 0.89)
- Occluded Pedestrian Detection Through Guided Attention in CNNs (CVPR 2018 · similarity 0.89)
- Bi-box Regression for Pedestrian Detection and Occlusion Estimation (ECCV 2018 · similarity 0.89)
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · similarity 0.88)
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.88)
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.
