Research map / Image generation / Image restoration
Iterative Crowd Counting
Viresh Ranjan, Hieu Le, Minh Hoai
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 low-res prediction and features from LR-CNN to the HR-CNN (full ic-CNN vs HR-CNN alone) (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 136.2 -> 69.8 | Improved |
| add the low-res density map prediction as input to HR-CNN (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 75.1 -> 69.8 | Improved |
| share LR-CNN convolutional features with HR-CNN (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 77.4 -> 69.8 | Improved |
| add HR-CNN on top of LR-CNN (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 78.5 -> 69.8 | Improved |
| use 1/4 resolution for the LR-CNN output instead of 1/8 (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 74.9 -> 69.8 | Improved |
| use 1/4 resolution for the LR-CNN output instead of full resolution (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 74.4 -> 69.8 | Improved |
| raise lambda_h from 1e-4 to 1e2 (ablation) | ic-CNN HR-CNN · ShanghaiTech Part A | MAE: 78.8 -> 69.8 | Improved |
| raise lambda_h from 1e2 to 1e4 (ablation) | ic-CNN HR-CNN · ShanghaiTech Part A | MAE: 69.8 -> 74.4 | Hurt |
| add a second stage to ic-CNN (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 69.8 -> 68.5 | Improved |
| add a third stage to ic-CNN (ablation) | ic-CNN · ShanghaiTech Part A | MAE: 69.8 -> 69.4 | No clear effect |
| add a second stage to ic-CNN (ablation) | ic-CNN · ShanghaiTech Part B | MAE: 10.4 -> 10.7 | Hurt |
Closest papers by meaning
- Adaptive Density Map Generation for Crowd Counting (ICCV 2019 · similarity 0.93)
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · similarity 0.93)
- Kernel-Based Density Map Generation for Dense Object Counting (TPAMI 2020 · similarity 0.92)
- 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)
- Crowd Counting and Density Estimation by Trellis Encoder-Decoder Networks (CVPR 2019 · similarity 0.91)
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.
