Research map / Image generation / Autoregressive and latent generation
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank
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 ImageNet pre-training (ablation) | VGG-16 crowd-counting CNN · UCF_CC_50 five-fold cross-validation | Ave MAE: 454.45 -> 328.54 | Helped |
| Add multi-scale sampling (ablation) | Pre-trained VGG-16 crowd-counting CNN · UCF_CC_50 five-fold cross-validation | Ave MAE: 328.54 -> 308.06 | Helped |
| Pre-train on ranked images, then fine-tune for counting (ablation) | Pre-trained, multi-scale VGG-16 · UCF_CC_50 five-fold cross-validation | Ave MAE: 308.06 -> 325.73 | Hurt |
| Alternate ranking and counting training (ablation) | Pre-trained, multi-scale VGG-16 · UCF_CC_50 five-fold cross-validation | Ave MAE: 308.06 -> 285.44 | Helped |
| Train ranking and counting simultaneously (ablation) | Pre-trained, multi-scale VGG-16 · UCF_CC_50 five-fold cross-validation | Ave MAE: 308.06 -> 279.60 | Helped |
| Add ranking data to counting training | Counting model trained on ShanghaiTech Part A · UCF_CC_50 transfer | MAE: 349.5 -> 337.6 | Helped |
| Add ranking data to counting training | Counting model trained on ShanghaiTech Part A · UCF_CC_50 transfer | MSE: 475.7 -> 434.3 | Helped |
| Use multi-task learning with keyword-collected unlabeled data | VGG-16 crowd-counting model · UCF_CC_50 | MAE: 295.8 -> 279.6 | Helped |
| Use multi-task learning with keyword-collected unlabeled data | VGG-16 crowd-counting model · UCF_CC_50 | MSE: 320.9 -> 388.9 | Hurt |
| Use query-by-example instead of keyword-collected unlabeled data | Multi-task crowd-counting model · UCF_CC_50 | MAE: 279.6 -> 291.5 | Hurt |
| Use multi-task learning with query-by-example unlabeled data | VGG-16 crowd-counting model · ShanghaiTech Part A | MAE: 73.6 -> 72.0 | Helped |
| Use multi-task learning with keyword-collected unlabeled data | VGG-16 crowd-counting model · ShanghaiTech Part B | MAE: 20.1 -> 13.7 | Helped |
Closest papers by meaning
- CrowdCLIP: Unsupervised Crowd Counting via Vision-Language Model (CVPR 2023 · similarity 0.91)
- Almost Unsupervised Learning for Dense Crowd Counting (AAAI 2019 · similarity 0.91)
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · similarity 0.89)
- Switching Convolutional Neural Network for Crowd Counting (CVPR 2017 · similarity 0.89)
- Single Domain Generalization for Crowd Counting (CVPR 2024 · similarity 0.88)
- Taste More, Taste Better: Diverse Data and Strong Model Boost Semi-Supervised Crowd Counting (CVPR 2025 · 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.