Research map / Learning theory and algorithms / Optimal transport
Distribution Matching for Crowd Counting
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 |
|---|---|---|---|
| replace Bayesian loss with DM-Count OT loss | VGG-19 counter · NWPU test | MAE: 105.4 to 88.4 | Helped |
| replace Bayesian loss with DM-Count OT loss | VGG-19 counter · UCF-QNRF | MAE: 88.7 to 85.6 | Helped |
| replace Gaussian-smoothed pixel-wise loss with DM-Count | VGG-19 counter · UCF-QNRF | MAE: 106.8 -> 85.6 | Helped |
| DM-Count vs CAN | crowd counter · NWPU test | RMSE: 386.5 vs 388.6 | No clear effect |
| add OT loss to counting + TV loss (ablation) | DM-Count · UCF-QNRF | MAE: 94.9 -> 85.6 | Helped |
| add TV loss to counting + OT loss (ablation) | DM-Count · UCF-QNRF | MAE: 89.3 -> 85.6 | Helped |
| increase Sinkhorn iterations 50 -> 100 (ablation) | DM-Count · UCF-QNRF | MAE: 90.8 -> 85.6 | Helped |
| increase Sinkhorn iterations 100 -> 120 (ablation) | DM-Count · UCF-QNRF | MAE: 85.6 -> 85.5 | No clear effect |
| increase TV weight lambda2 from 0.01 to 0.1 (ablation) | DM-Count · UCF-QNRF | MAE: 85.6, 87.8 to 88.5 | Hurt |
| increase OT weight lambda1 from 0.01 to 0.1 (ablation) | DM-Count · UCF-QNRF | MAE: 87.2, 86.2 to 85.6 | Helped |
| DM-Count instead of Bayesian loss under noisy annotations | VGG-19 counter · UCF-QNRF noisy | MAE: 108.4 -> 105.6 | Helped |
| DM-Count instead of Bayesian loss | VGG-19 counter · UCF-QNRF | PSNR: 34.55 -> 40.65 | Helped |
Closest papers by meaning
- Modeling Noisy Annotations for Crowd Counting (NeurIPS 2020 · similarity 0.91)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · similarity 0.89)
- Counting without Scale: Scale-Consistent Error Correction for Crowd Counting (NeurIPS 2026 · similarity 0.89)
- Learning to Count via Unbalanced Optimal Transport (AAAI 2021 · similarity 0.88)
- Single Domain Generalization for Crowd Counting (CVPR 2024 · similarity 0.88)
- Gaussian Density Splatting Network (NeurIPS 2026 · similarity 0.87)
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.