Research map / Point clouds / Registration and matching
Rethinking Counting and Localization in Crowds: A Purely Point-Based Framework
Qingyu Song, Changan Wang, Zhengkai Jiang, Yabiao Wang, Ying Tai, Chengjie Wang, Jilin Li, Feiyue Huang, Yang Wu
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 |
|---|---|---|---|
| use grid layout of reference points instead of center layout (ablation) | P2PNet · ShanghaiTech PartA | MAE: 53.7 -> 52.74 | Improved |
| use grid layout of reference points instead of center layout (ablation) | P2PNet · ShanghaiTech PartA | nAP (delta 0.05:0.05:0.50): 61.7 -> 64.4 | Improved |
| use stride 8 feature map instead of stride 4 (ablation) | P2PNet · ShanghaiTech PartA | MAE: 53.51 -> 52.74 | Improved |
| use stride 8 feature map instead of stride 16 (ablation) | P2PNet · ShanghaiTech PartA | MAE: 54.3 -> 52.74 | Improved |
| use finer stride 4 instead of stride 8 (ablation) | P2PNet · ShanghaiTech PartA | nAP (delta 0.05:0.05:0.50): 64.4 -> 66.8 | Improved |
| use stride 16 instead of stride 8 (ablation) | P2PNet · ShanghaiTech PartA | nAP (delta 0.05:0.05:0.50): 64.4 -> 52.4 | Hurt |
| increase reference points K from 1 to 4 (ablation) | P2PNet · ShanghaiTech PartA | MAE: 54.08 -> 52.74 | Improved |
| increase reference points K from 4 to 8 (ablation) | P2PNet · ShanghaiTech PartA | MAE: 52.74 -> 53.43 | Hurt |
| increase reference points K from 4 to 12 (ablation) | P2PNet · ShanghaiTech PartA | MAE: 52.74 -> 54.13 | Hurt |
| use only K=1 reference point (ablation) | P2PNet · ShanghaiTech PartA | MAE: 54.08 | No clear effect |
| use P2PNet instead of ADSCNet (second best) on ShanghaiTech PartA | ADSCNet · ShanghaiTech PartA | MAE: 55.4 -> 52.74 | Improved |
| use P2PNet instead of ADSCNet on UCF-QNRF | ADSCNet · UCF-QNRF | MAE: 71.3 -> 85.32 | Hurt |
Closest papers by meaning
- Point-to-Region Loss for Semi-Supervised Point-Based Crowd Counting (CVPR 2025 · similarity 0.92)
- Recurrent Attentive Zooming for Joint Crowd Counting and Precise Localization (CVPR 2019 · similarity 0.92)
- Optimal Transport Minimization: Crowd Localization on Density Maps for Semi-Supervised Counting (CVPR 2023 · similarity 0.92)
- AutoScale: Learning to Scale for Crowd Counting (IJCV 2022 · similarity 0.91)
- Robust Zero-Shot Crowd Counting and Localization with Adaptive Resolution SAM (ECCV 2024 · similarity 0.91)
- Learning to Count via Unbalanced Optimal Transport (AAAI 2021 · similarity 0.90)
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.
