Research map / Detection and segmentation / Open-set and few-shot detection
Generalizable Pedestrian Detection: The Elephant in the Room
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a general object detector rather than the pedestrian-specific BGCNet design | Cascade R-CNN vs BGCNet with HRNet · Caltech→CityPersons | log-average miss rate: 41.4 -> 36.5 | Helped |
| Use a general object detector rather than CSP | Cascade R-CNN vs CSP with HRNet · ECP→ECP Reasonable | log-average miss rate: 19.4 -> 6.9 | Helped |
| Use a general object detector rather than CSP | Cascade R-CNN vs CSP with HRNet · ECP→CityPersons Reasonable | log-average miss rate: 11.5 -> 10.9 | Helped |
| Replace the pedestrian-specific FRCNN adaptations with vanilla FRCNN (ablation) | VGG-16 FRCNN · CityPersons→Caltech | log-average miss rate: 17.8 -> 17.6 | Helped |
| Pretrain Cascade R-CNN on CrowdHuman | Cascade R-CNN with HRNet · Caltech Reasonable | log-average miss rate | Helped |
| Train on CrowdHuman rather than CityPersons | Cascade R-CNN · ECP Reasonable | log-average miss rate: 6.9 -> 17.9 | Hurt |
| Progressively fine-tune from Wider Pedestrian to ECP instead of training on ECP alone (ablation) | Cascade R-CNN with HRNet · CityPersons Reasonable | log-average miss rate: 10.9 -> 9.7 | Helped |
| Progressively fine-tune from CrowdHuman to ECP instead of training on ECP alone (ablation) | Cascade R-CNN with HRNet · CityPersons Reasonable | log-average miss rate: 10.9 -> 10.3 | Helped |
| Progressively fine-tune on ECP after pretraining on Wider Pedestrian and CrowdHuman rather than merging all three datasets (ablation) | Cascade R-CNN with HRNet · CityPersons Reasonable | log-average miss rate: 10.9 -> 9.7 | Helped |
| Use ECP alone rather than progressive pretraining on Wider Pedestrian followed by ECP (ablation) | MobileNet Cascade R-CNN · CityPersons Reasonable | log-average miss rate: 11.4 -> 19.1 | Hurt |
| Progressively fine-tune from CrowdHuman to ECP (ablation) | MobileNet Cascade R-CNN · CityPersons Reasonable | log-average miss rate: 12.0 -> 11.9 | Helped |
| Progressively fine-tune from Wider Pedestrian to ECP (ablation) | MobileNet Cascade R-CNN · CityPersons Reasonable | log-average miss rate: 12.0 -> 11.4 | Helped |
Closest papers by meaning
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · similarity 0.90)
- High-Level Semantic Feature Detection: A New Perspective for Pedestrian Detection (CVPR 2019 · similarity 0.88)
- WILDTRACK: A Multi-Camera HD Dataset for Dense Unscripted Pedestrian Detection (CVPR 2018 · similarity 0.88)
- Localized Semantic Feature Mixers for Efficient Pedestrian Detection in Autonomous Driving (CVPR 2023 · similarity 0.88)
- Multispectral Pedestrian Detection with Sparsely Annotated Label (AAAI 2025 · similarity 0.85)
- Causal Mode Multiplexer: A Novel Framework for Unbiased Multispectral Pedestrian Detection (CVPR 2024 · 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.