Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain-General Crowd Counting in Unseen Scenarios
Read the paperCheck 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 the proposed domain-general method instead of DG-MAN | Crowd counting trained on SHA or SHB · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MAE: 17.3 -> 12.6; 129.1 -> 119.4; 130.7 -> 121.8; 182.4 -> 179.1 | Helped |
| Use the proposed domain-general method instead of DG-MAN | Crowd counting trained on SHA or SHB · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MSE: 28.7 -> 24.6; 238.2 -> 216.6; 225.1 -> 203.1; 325.8 -> 316.2 | Helped |
| Add MLDG meta-learning to the baseline (ablation) | Baseline crowd counter · SHA→SHB | MAE: 22.0 -> 17.1 | Helped |
| Add the domain-invariant branch and its reconstruction loss (ablation) | +MLDG · SHA→SHB | MAE: 17.1 -> 13.6 | Helped |
| Add the domain-specific branch and orthogonal loss (ablation) | +DI-B · SHA→SHB | MAE: 13.6 -> 12.6 | Helped |
| Add feature reconstruction loss (ablation) | Full method · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MAE: 15.4 -> 12.6; 132.8 -> 119.4; 175.6 -> 121.8; 266.9 -> 179.1 | Helped |
| Add feature orthogonal loss (ablation) | Full method · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MAE: 17.8 -> 12.6; 132.2 -> 119.4; 148.6 -> 121.8; 219.8 -> 179.1 | Helped |
| Add hard predicted region reinforcement (ablation) | Full method · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MAE: 13.4 -> 12.6; 124.0 -> 119.4; 127.8 -> 121.8; 189.3 -> 179.1 | Helped |
| Dynamically refine sub-domain assignments rather than cluster once (ablation) | Full method · SHA→SHB, SHA→QNRF, SHB→SHA, SHB→QNRF | MAE: 14.0 -> 12.6; 125.6 -> 119.4; 128.3 -> 121.8; 198.8 -> 179.1 | Helped |
| Use base feature f for dynamic sub-domain division instead of z (ablation) | Full method · SHA→SHB | MAE: 15.4 -> 12.6 | Helped |
| Use re-encoded invariant feature tilde f for dynamic sub-domain division instead of z (ablation) | Full method · SHA→SHB | MAE: 21.4 -> 12.6 | Helped |
| Use average re-encoded domain-specific feature tilde z for clustering instead of z (ablation) | Full method · SHA→SHB | MAE: 16.5 -> 12.6 | Helped |
Closest papers by meaning
- Single Domain Generalization for Crowd Counting (CVPR 2024 · similarity 0.92)
- Explicit Invariant Feature Induced Cross-Domain Crowd Counting (AAAI 2023 · similarity 0.91)
- Error-Aware Density Isomorphism Reconstruction for Unsupervised Cross-Domain Crowd Counting (AAAI 2021 · similarity 0.90)
- Free Lunch Enhancements for Multi-modal Crowd Counting (CVPR 2025 · similarity 0.86)
- Gramformer: Learning Crowd Counting via Graph-Modulated Transformer (AAAI 2024 · similarity 0.86)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · 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.