Research map / Detection and segmentation / CNN features
Pedestrian Detection With Autoregressive Network Phases
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 |
|---|---|---|---|
| Add the R-CNN second stage to AR-RPN | AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 6.45 | Helped |
| Add a fourth autoregressive phase (ablation) | AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 8.68 | Hurt |
| Use small channel widths instead of medium (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 8.62 | Hurt |
| Use large channel widths instead of medium (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 8.33 | Hurt |
| Replace learned convolutional resampling with separate bilinear resampling and convolution (ablation) | AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01% -> 9.45% | Hurt |
| Use only 2 phases instead of 3 (ablation) | AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 8.32 | Hurt |
| Use only 1 phase instead of 3 (ablation) | AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 10.16 | Hurt |
| Remove intermediate predictions and their concatenation between phases (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.01 -> 9.06 | Hurt |
| Use lenient-to-strict labels rather than strict-to-lenient labels (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 9.03 -> 8.01 | Helped |
| Use lenient-to-strict labels rather than moderate-to-moderate labels (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.94 -> 8.01 | Helped |
| Use lenient-to-strict labels rather than strict-to-strict labels (ablation) | 3-phase AR-RPN · Caltech Reasonable (original annotations) | MR^-2: 8.43 -> 8.01 | Helped |
| Use convolutional resampling rather than two-step resampling (ablation) | AR-RPN · Caltech Reasonable | GPU memory usage: 41% less | Helped |
Closest papers by meaning
- Where, What, Whether: Multi-Modal Learning Meets Pedestrian Detection (CVPR 2020 · similarity 0.88)
- What Can Help Pedestrian Detection? (CVPR 2017 · similarity 0.87)
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · similarity 0.87)
- CityPersons: A Diverse Dataset for Pedestrian Detection (CVPR 2017 · similarity 0.86)
- Multi-Rate Gated Recurrent Convolutional Networks for Video-Based Pedestrian Re-Identification (AAAI 2018 · similarity 0.86)
- Generalizable Pedestrian Detection: The Elephant in the Room (CVPR 2021 · similarity 0.86)
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.