Research map / Label-efficient and robust learning / Domain adaptation and generalization
Variational Attention: Propagating Domain-Specific Knowledge for Multi-Domain Learning in Crowd Counting
Binghui Chen, Zhaoyi Yan, Ke Li, Pengyu Li, Biao Wang, Wangmeng Zuo, Lei Zhang
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 |
|---|---|---|---|
| merge all datasets for joint training (JT) instead of individual training | HRNet baseline · SHB (3-Joint) | MAE: 8.8 -> 9.3 | Hurt |
| merge four datasets for joint training (JT) | HRNet baseline · SHB (4-Joint) | MAE: 8.8 -> 9.7 | Hurt |
| add SE attention to joint training | HRNet baseline · SHB (3-Joint) | MAE: 8.8 -> 9.4 | Hurt |
| add VA (variational attention with Gaussian mixture latent) (ablation) | JT baseline · SHA/SHB/QNRF (3-Joint) | MAE: 60.2 -> 57.5; 9.3 -> 7.6; 92.8 -> 87.9 | Improved |
| add clustering labels (InVA c=3) on top of VA (ablation) | VA · SHA/SHB/QNRF (3-Joint) | MAE: 57.5 -> 57.3; 7.6 -> 7.5; 87.9 -> 86.7 | Improved |
| add sub-Gaussian components (k=3) on top of InVA c=3 (ablation) | InVA · SHA/SHB/QNRF (3-Joint) | MAE: 57.3 -> 56.7; 7.5 -> 6.9; 86.7 -> 85.2 | Improved |
| add distribution regularizer L_reg (ablation) | DKPNet 4-Joint · NWPU val (4-Joint) | MAE: 64.2 -> 61.8 | Improved |
| add distribution regularizer L_reg (ablation) | DKPNet 4-Joint · QNRF (4-Joint) | MAE: 86.9 -> 81.4 | Improved |
| replace SE attention with DKPNet domain-specific attention (ablation) | SE model · NWPU val (4-Joint) | MAE: 66.4 -> 61.8 | Improved |
| increase number of clusters c from 3 to 4 (ablation) | InVA 3-Joint · SHA (3-Joint) | MAE: 57.3 -> 59.6 | Hurt |
| increase number of clusters c from 3 to 2 (ablation) | InVA 3-Joint · SHB (3-Joint) | MAE: 7.5 -> 7.9 | Hurt |
| increase sub-Gaussians k from 3 to 4 (ablation) | InVA c=3 · SHA (3-Joint) | MAE: 56.7 -> 57.7 | Hurt |
Closest papers by meaning
- Domain-General Crowd Counting in Unseen Scenarios (AAAI 2023 · similarity 0.89)
- Explicit Invariant Feature Induced Cross-Domain Crowd Counting (AAAI 2023 · similarity 0.88)
- Boosting Crowd Counting via Multifaceted Attention (CVPR 2022 · similarity 0.88)
- Attention Scaling for Crowd Counting (CVPR 2020 · similarity 0.87)
- Gramformer: Learning Crowd Counting via Graph-Modulated Transformer (AAAI 2024 · similarity 0.87)
- Free Lunch Enhancements for Multi-modal Crowd Counting (CVPR 2025 · 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.
