Research map / Detection and segmentation / Open-set and few-shot detection
EuroCity Persons: A Novel Benchmark for Person Detection in Traffic Scenes
Markus Braun, Sebastian Krebs, Fabian B. Flohr, Dariu M. Gavrila
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 |
|---|---|---|---|
| pre-train on ECP then fine-tune on KITTI instead of KITTI only | Faster R-CNN · KITTI validation (moderate) | AP: 72.3 -> 80.8 | Improved |
| pre-train on CP instead of ECP before KITTI fine-tuning | Faster R-CNN · KITTI validation (moderate) | AP: 80.8 -> 77.5 | Hurt |
| pre-train on ECP then fine-tune on CityPersons instead of CP only | Faster R-CNN · CityPersons validation (reasonable) | LAMR: 17.2 -> 14.9 | Improved |
| pre-train on KITTI before CityPersons fine-tuning | Faster R-CNN · CityPersons validation (reasonable) | LAMR: 17.2 -> 17.2 | No clear effect |
| train only on ECP without fine-tuning on CityPersons | Faster R-CNN · CityPersons validation (reasonable) | LAMR: 17.2 -> 22.5 | Hurt |
| train on up to 80% occlusion and all heights instead of only <=40% occlusion, no upscaling | Faster R-CNN · ECP test (occluded) | LAMR: 55.1 -> 33.9 | Improved |
| upscale images by 1.3 (bilinear) (ablation) | Faster R-CNN all · ECP test (reasonable) | LAMR: 9.4 -> 8.1 | Improved |
| enable OHEM (ablation) | R-FCN · ECP test (occluded) | LAMR: 45.6 -> 44.0 | Improved |
| add ignore-region handling (ablation) | Faster R-CNN baseline · ECP test (reasonable) | LAMR: 10.9 -> 9.4 | Improved |
| add Biternion orientation multitask loss (ablation) | Faster R-CNN baseline · ECP test (reasonable) | LAMR: 9.4 -> 10.1 | Hurt |
| add rider class training, with pedestrian class enforced (ablation) | Faster R-CNN baseline · ECP test pedestrians (enforce) | LAMR: 11.0 -> 10.3 | Improved |
| train on day and night rather than night only, testing on night (ablation) | Faster R-CNN · ECP night test (reasonable) | LAMR: 18.6 -> 23.1 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Faster R-CNN_small — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 1 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EuroCity Persons 2.0: A Large and Diverse Dataset of Persons in Traffic (TPAMI 2024 · similarity 0.92)
- CityPersons: A Diverse Dataset for Pedestrian Detection (CVPR 2017 · similarity 0.91)
- Towards Reaching Human Performance in Pedestrian Detection (TPAMI 2017 · similarity 0.89)
- Generalizable Pedestrian Detection: The Elephant in the Room (CVPR 2021 · similarity 0.89)
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · similarity 0.89)
- Artificial Dummies for Urban Dataset Augmentation (AAAI 2021 · 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.
