Research map / 3D vision / Depth and stereo
Switching Convolutional Neural Network for Crowd Counting
Check 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 |
|---|---|---|---|
| Route image patches to independently trained regressors with different receptive fields | Switch-CNN · ShanghaiTech Part A and Part B | MAE: 110.2 -> 90.4; 26.4 -> 21.6 | Helped |
| Use three distinct regressors rather than a two-regressor subset (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 98.87 -> 90.41; 110.88 -> 90.41; 126.65 -> 90.41 | Helped |
| Use only R1 and R3 instead of all three regressors (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 98.87 | Hurt |
| Use only R1 and R2 instead of all three regressors (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 110.88 | Hurt |
| Use only R2 and R3 instead of all three regressors (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 126.65 | Hurt |
| Replace learned differential patch grouping with manual count-based grouping (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 99.56 | Hurt |
| Replace learned differential patch grouping with manual inter-head-distance grouping (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 94.93 | Hurt |
| Train the switch alone with fixed labels instead of coupled switch-regressor training (ablation) | Switch-CNN · ShanghaiTech Part A | MAE: 90.41 -> 103.26 | Hurt |
| Use three distinct regressors rather than a two-regressor subset (ablation) | Switch-CNN · UCF_CC_50 | MAE: 369.58 -> 318.07; 362.22 -> 318.07; 334.66 -> 318.07 | Mixed |
| Replace learned differential grouping with manual count-based grouping (ablation) | Switch-CNN · UCF_CC_50 | MAE: 318.07 -> 319.16 | No clear effect |
| Replace learned differential grouping with manual inter-head-distance grouping (ablation) | Switch-CNN · UCF_CC_50 | MAE: 318.07 -> 358.78 | Hurt |
| Use Switch-CNN instead of MCNN | Crowd counting · UCF_CC_50 | MSE: 425.26 -> 439.2 | Hurt |
Closest papers by meaning
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · similarity 0.93)
- Top-Down Feedback for Crowd Counting Convolutional Neural Network (AAAI 2018 · similarity 0.92)
- Context-Aware Crowd Counting (CVPR 2019 · similarity 0.91)
- Divide and Grow: Capturing Huge Diversity in Crowd Images With Incrementally Growing CNN (CVPR 2018 · similarity 0.91)
- Wide-Area Crowd Counting via Ground-Plane Density Maps and Multi-View Fusion CNNs (CVPR 2019 · similarity 0.91)
- Gramformer: Learning Crowd Counting via Graph-Modulated Transformer (AAAI 2024 · similarity 0.88)
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.