Research map / Label-efficient and robust learning / Recognition and long-tail
Weakly-Supervised Crowd Counting Learns from Sorting rather than Locations
Yifan Yang, Guorong Li, Zhe Wu, Li Su, Qingming Huang, Nicu Sebe
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 the soft-sort network (sorting) to the counting network (ablation) | regression network · WorldExpo10 | MAE: 20.1 -> 9.6 | Improved |
| add the regression network to the sorting network (ablation) | sorting network · WorldExpo10 | sorting accuracy (%): 50.6 -> 78.2 | Improved |
| use Rayleigh soft-labels instead of hard one-hot labels (ablation) | sorting network · WorldExpo10 | MAE: 13.2 -> 9.6 | Improved |
| use Rayleigh soft-labels instead of hard one-hot labels (ablation) | sorting network · WorldExpo10 | sorting accuracy (%): 89.1 -> 78.2 | Hurt |
| replace VGG backbone with R3D 3D-conv backbone (ablation) | full method · WorldExpo10 | MAE: 9.6 -> 10.1 | Hurt |
| use max and average pooling instead of only average pooling (ablation) | pyramidal pooling · WorldExpo10 | MAE: 11.4 -> 9.6 | Improved |
| sort 4 frames instead of 3 (ablation) | sorting network · WorldExpo10 | MAE: 9.6 -> 12.5 | Hurt |
| sort 5 frames instead of 3 (ablation) | sorting network · WorldExpo10 | MAE: 9.6 -> 13.1 | Hurt |
| sort 4 frames instead of 3 (ablation) | sorting network · WorldExpo10 | sorting accuracy (%): 78.2 -> 30.1 | Hurt |
| sort 5 frames instead of 3 (ablation) | sorting network · WorldExpo10 | sorting accuracy (%): 78.2 -> 16.3 | Hurt |
| sort 5 frames instead of no sorting network (ablation) | regression network · WorldExpo10 | MAE: 20.1 -> 13.1 | Improved |
Closest papers by meaning
- Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (CVPR 2018 · similarity 0.91)
- Active Crowd Counting with Limited Supervision (ECCV 2020 · similarity 0.90)
- Semi-Supervised Crowd Counting via Self-Training on Surrogate Tasks (ECCV 2020 · similarity 0.90)
- Crowd Counting and Density Estimation by Trellis Encoder-Decoder Networks (CVPR 2019 · similarity 0.90)
- Iterative Crowd Counting (ECCV 2018 · similarity 0.90)
- Completely Self-Supervised Crowd Counting via Distribution Matching (ECCV 2022 · similarity 0.90)
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.
