Research map / Probabilistic and scientific modeling / Variational and generative sampling
Modeling Noisy Annotations 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 L2 loss with noisy-annotation loss | VGG19 density estimator · UCF-QNRF | MAE: 98.7 -> 85.8 | Helped |
| replace L2 loss with noisy-annotation loss | CSRNet · UCF-QNRF | MAE: 110.6 -> 96.5 | Helped |
| replace L2 loss with noisy-annotation loss | MCNN · UCF-QNRF | MAE: 186.4 -> 177.4 | Helped |
| replace Bayesian loss with noisy-annotation loss | VGG19 · NWPU-Crowd | MSE: 454.2 -> 534.2 | Hurt |
| use full covariance instead of diagonal covariance (ablation) | noisy-annotation loss · UCF-QNRF | MAE: 103.2 -> 92.6 | Helped |
| add regularization term to FullCov (ablation) | noisy-annotation loss · UCF-QNRF | MAE: 92.6 -> 89.9 | Helped |
| add background (virtual dot) model (ablation) | FullCov + Reg · UCF-QNRF | MAE: 89.9 -> 85.8 | Helped |
| Gaussian instead of Gamma approximation of marginal (ablation) | noisy-annotation loss · UCF-QNRF | MAE: 89.7 -> 85.8 | Helped |
| increase preserved variance of low-rank covariance above 60% (ablation) | noisy-annotation loss · Mall | MAE | Hurt |
| increase annotation noise up to 64 pixels (ablation) | noisy-annotation loss vs L2 and BL · UCF-QNRF noisy annotations | MAE | Helped |
| vary density map variance beta (ablation) | noisy-annotation loss · UCF-QNRF | MAE | No clear effect |
| set annotation variance alpha >= 8 instead of very small (ablation) | noisy-annotation loss · UCF-QNRF | MAE | Helped |
Closest papers by meaning
- Distribution Matching for Crowd Counting (NeurIPS 2020 · similarity 0.91)
- Counting without Scale: Scale-Consistent Error Correction for Crowd Counting (NeurIPS 2026 · similarity 0.89)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · similarity 0.89)
- Crowd Counting in the Frequency Domain (CVPR 2022 · similarity 0.88)
- Single Domain Generalization for Crowd Counting (CVPR 2024 · similarity 0.87)
- Taste More, Taste Better: Diverse Data and Strong Model Boost Semi-Supervised 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.