Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Crowd Counting with Decomposed Uncertainty
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 |
|---|---|---|---|
| Add bootstrap ensemble and aleatoric uncertainty outputs (ablation) | DUBNet · UCF-50 | MAE: 258.4 -> 243.8 | Helped |
| Add bootstrap ensemble and aleatoric uncertainty outputs (ablation) | DUBNet · UCF-QNRF | MAE: 120.9 -> 105.6 | Helped |
| Add DUB extension (ablation) | CSRNet · UCF-50 | MAE: 266.1 -> 244.9 | Helped |
| Add DUB extension (ablation) | CSRNet · UCF-QNRF | MAE: 135.5 -> 127.1 | Helped |
| Add DUB extension (ablation) | MCNN · UCF-50 | MAE: 377.6 -> 359.4 | Helped |
| Add DUB extension (ablation) | MCNN · UCF-QNRF | MAE: 277.0 -> 254.6 | Helped |
| Use shared-backbone DUBNet rather than independent full-bootstrap networks (ablation) | Bootstrap crowd counting model · ShanghaiTech Part A test | Predictive variance: 47.6 -> 45.8 | No clear effect |
| Use shared-backbone DUBNet rather than independent full-bootstrap networks (ablation) | Bootstrap crowd counting model · ShanghaiTech Part B test | Predictive variance: 1.17 -> 1.02 | No clear effect |
| Apply DUB extension to the final output layer | DUBNet · ShanghaiTech Part A test | Inference runtime on CPU: +2% | Hurt |
| Apply DUB extension to the final output layer | DUBNet · ShanghaiTech Part A test | Inference runtime on GPU: 0.5% | Hurt |
| Use DUBNet instead of best reported baseline | Crowd counting models · ShanghaiTech A, ShanghaiTech B, UCF-50, and UCF-QNRF | MAE and RMSE | Mixed |
Closest papers by meaning
- Counting without Scale: Scale-Consistent Error Correction for Crowd Counting (NeurIPS 2026 · similarity 0.87)
- Crowd Counting via Adversarial Cross-Scale Consistency Pursuit (CVPR 2018 · similarity 0.87)
- Residual Regression With Semantic Prior for Crowd Counting (CVPR 2019 · similarity 0.87)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · similarity 0.85)
- Single Domain Generalization for Crowd Counting (CVPR 2024 · similarity 0.84)
- Free Lunch Enhancements for Multi-modal Crowd Counting (CVPR 2025 · similarity 0.84)
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.