Research map / Detection and segmentation / CNN features
CityPersons: A Diverse Dataset for Pedestrian Detection
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 quantile-based RPN scales instead of sparse default scales (ablation) | FasterRCNN on Caltech · Caltech | MR^O: 20.98 -> 18.11 | Helped |
| Upscale input images (ablation) | FasterRCNN on Caltech · Caltech | MR^O: 18.11 -> 14.37; + 3.74 | Helped |
| Switch from Caffe SGD to Adam (ablation) | FasterRCNN on Caltech · Caltech | MR^O: 14.37 -> 12.70 | Helped |
| Avoid sampling ignore regions during RPN training (ablation) | FasterRCNN on Caltech · Caltech | MR^O: 12.70 -> 11.37; 1.33 MR pp | Helped |
| Reduce feature stride by removing VGG16's fourth max-pooling layer (ablation) | FasterRCNN on Caltech · Caltech | MR^O: 11.37 -> 10.27 | Helped |
| Pre-train on CityPersons, then fine-tune on Caltech | FasterRCNN · Caltech test, reasonable, MR^N | MR^N: 5.8 -> 5.1; + 0.7 | Helped |
| Pre-train on CityPersons, then fine-tune on Caltech | FasterRCNN · Caltech test, reasonable, IoU 0.75 | MR^N: 30.6 -> 25.8; + 4.8 | Helped |
| Pre-train on CityPersons, then fine-tune on KITTI | FasterRCNN · KITTI validation, reasonable, IoU 0.5 | MR: 8.4 -> 5.9; + 2.5 | Helped |
| Add predicted semantic channels alongside RGB | FasterRCNN on CityPersons · CityPersons validation, scale [50, ∞] | MR: 15.4 -> 14.8; + 0.6 | Helped |
| Add predicted semantic channels alongside RGB | FasterRCNN on CityPersons · CityPersons validation, scale [100, ∞] | MR: 7.9 -> 8.0; + 0.1 | Hurt |
| Add predicted semantic channels alongside RGB | FasterRCNN on CityPersons · CityPersons validation, scale [50, 75] | MR: 25.6 -> 22.6; + 3.0 | Helped |
| Add ignore-region annotations to segment-box training labels (ablation) | FasterRCNN on CityPersons · CityPersons validation | MR: 22.54 -> 21.31; + 1.23 | Helped |
Closest papers by meaning
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · similarity 0.90)
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.88)
- MixedPeds: Pedestrian Detection in Unannotated Videos Using Synthetically Generated Human-Agents for Training (AAAI 2018 · similarity 0.88)
- Artificial Dummies for Urban Dataset Augmentation (AAAI 2021 · similarity 0.87)
- Occluded Pedestrian Detection Through Guided Attention in CNNs (CVPR 2018 · similarity 0.87)
- Expecting the Unexpected: Training Detectors for Unusual Pedestrians With Adversarial Imposters (CVPR 2017 · similarity 0.87)
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.