Research map / Learning theory and algorithms / Optimal transport
Towards a Universal Model for Cross-Dataset Crowd Counting
Zhiheng Ma, Xiaopeng Hong, Xing Wei, Yunfeng Qiu, Yihong Gong
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 scale distribution alignment (SDA) to jointly trained BL | BL · UCF-QNRF (joint multi-dataset training) | MAE: 97.3 -> 79.2 | Improved |
| add SDA to jointly trained BL | BL · Shanghai A (joint multi-dataset training) | MAE: 66.1 -> 53.6 | Improved |
| add SDA to jointly trained BL | BL · UCF CC 50 (joint multi-dataset training) | MAE: 231.1 -> 169.4 | Improved |
| add SDA to jointly trained CSRNet | CSRNet · UCF-QNRF (joint multi-dataset training) | MAE: 158.0 -> 96.3 | Improved |
| add SDA to jointly trained M-SFANet | M-SFANet · UCF-QNRF (joint multi-dataset training) | MAE: 111.8 -> 79.5 | Improved |
| add SDA to jointly trained DM | DM · UCF-QNRF (joint multi-dataset training) | MAE: 102.6 -> 80.7 | Improved |
| train jointly on multiple datasets without alignment | BL · UCF-QNRF | MAE: 88.7 -> 97.3 | Hurt |
| train jointly on multiple datasets without alignment | M-SFANet · UCF CC 50 | MAE: 162.3 -> 233.6 | Hurt |
| add SDA to BL trained on each dataset separately (ablation) | BL · UCF-QNRF (single-dataset training) | MAE: 88.7 -> 83.3 | Improved |
| add multi-dataset training to SDA+BL (ablation) | SDA+BL · UCF-QNRF | MAE: 83.3 -> 79.2 | Improved |
| align with ground-truth scale distributions (upper bound) (ablation) | BL · UCF CC 50 | MAE: 229.3 -> 114.7 | Improved |
| align with ground-truth scale distributions (upper bound) (ablation) | BL · UCF-QNRF | MAE: 88.7 -> 66.2 | Improved |
Closest papers by meaning
- Counting without Scale: Scale-Consistent Error Correction for Crowd Counting (NeurIPS 2026 · similarity 0.90)
- Crowd Counting via Adversarial Cross-Scale Consistency Pursuit (CVPR 2018 · similarity 0.89)
- Context-Aware Crowd Counting (CVPR 2019 · similarity 0.89)
- Cross-View Cross-Scene Multi-View Crowd Counting (CVPR 2021 · similarity 0.88)
- Crowd Counting With Deep Structured Scale Integration Network (ICCV 2019 · similarity 0.88)
- Crowd Counting With Partial Annotations in an Image (ICCV 2021 · 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.
