Research map / Label-efficient and robust learning / Recognition and long-tail
Divide and Grow: Capturing Huge Diversity in Crowd Images With Incrementally Growing CNN
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 |
|---|---|---|---|
| use hierarchical differential training with 8 expert regressors and an expert classifier (ablation) | IG-CNN · ShanghaiTech Part A | MAE: 120.9 -> 72.5 | Helped |
| grow from 8 to 16 experts (ablation) | IG-CNN · ShanghaiTech Part A | MAE: 72.5 -> 74.6 | Hurt |
| grow to 8 rather than 7 experts using an unbalanced tree (ablation) | IG-CNN · ShanghaiTech Part A | MAE: 78.1 -> 72.5 | Helped |
| use the hierarchical expert model (ablation) | crowd counting · ShanghaiTech Part A | MAE: 75.1 -> 72.5 | Helped |
| use a softmax mixture-of-experts gating approach instead of hard specialist selection (ablation) | crowd counting · ShanghaiTech Part A | MAE: 72.5 -> 281.8 | Hurt |
| use IG-CNN instead of CP-CNN | crowd counting · ShanghaiTech Part A | MAE: 73.6 -> 72.5 | Helped |
| use IG-CNN instead of CP-CNN | crowd counting · ShanghaiTech Part B | MAE: 20.1 -> 13.6 | Helped |
| use IG-CNN instead of CP-CNN | crowd counting · UCF CC 50 | MAE: 295.8 -> 291.4 | Helped |
| use IG-CNN instead of CP-CNN | crowd counting · WorldExpo’10 Average | MAE: 8.9 -> 11.3 | Hurt |
Closest papers by meaning
- Top-Down Feedback for Crowd Counting Convolutional Neural Network (AAAI 2018 · similarity 0.92)
- Switching Convolutional Neural Network for Crowd Counting (CVPR 2017 · similarity 0.91)
- Context-Aware Crowd Counting (CVPR 2019 · similarity 0.91)
- Almost Unsupervised Learning for Dense Crowd Counting (AAAI 2019 · similarity 0.90)
- Gramformer: Learning Crowd Counting via Graph-Modulated Transformer (AAAI 2024 · similarity 0.87)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · 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.