Research map / Architectures and efficiency / Attention and transformers
Relational Attention Network for Crowd Counting
Anran Zhang, Jiayi Shen, Zehao Xiao, Fan Zhu, Xiantong Zhen, Xianbin Cao, Ling Shao
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 |
|---|---|---|---|
| add concatenation-based intra-relation fusion of LSA and GSA (ablation) | GSA w/o RM · ShanghaiTech A (ablation) | RMSE: 107.2 -> 103.4 | Improved |
| fuse LSA and GSA by element-wise sum instead of concatenation (ablation) | LSA & GSA · ShanghaiTech A (ablation) | RMSE: 103.4 -> 113.0 | Hurt |
| use concatenation rather than sum in intra-relation (ablation) | LSA & GSA · ShanghaiTech A (ablation) | MAE: 63.4 -> 62.2 | Improved |
| add inter-relation on top of intra-relation (concat) (ablation) | LSA & GSA · ShanghaiTech A (ablation) | MAE: 62.2 -> 59.4 | Improved |
| add inter-relation on top of intra-relation (concat) (ablation) | LSA & GSA · ShanghaiTech A (ablation) | RMSE: 103.4 -> 102.0 | Improved |
| add global self-attention alongside local self-attention with inter-relation (ablation) | LSA w/o RM · ShanghaiTech A (ablation) | MAE: 63.6 -> 59.4 | Improved |
| use GSA instead of LSA without relation module (ablation) | LSA w/o RM · ShanghaiTech A (ablation) | MAE: 63.6 -> 62.3 | Improved |
| use GSA instead of LSA without relation module (ablation) | LSA w/o RM · ShanghaiTech A (ablation) | RMSE: 113.7 -> 107.2 | Improved |
| replace best prior method SANet with RANet | SANet · ShanghaiTech A | MAE: 67.0 -> 59.4 | Improved |
| replace Idrees et al with RANet | Idrees et al · UCF-QNRF | MAE: 132 -> 111 | Improved |
Closest papers by meaning
- Attention Scaling for Crowd Counting (CVPR 2020 · similarity 0.91)
- Shallow Feature Based Dense Attention Network for Crowd Counting (AAAI 2020 · similarity 0.90)
- DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density Estimation (CVPR 2018 · similarity 0.89)
- Crowd Counting and Density Estimation by Trellis Encoder-Decoder Networks (CVPR 2019 · similarity 0.88)
- Learning Spatial Awareness to Improve Crowd Counting (ICCV 2019 · similarity 0.88)
- Boosting Crowd Counting via Multifaceted Attention (CVPR 2022 · 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.
