Research map / Architectures and efficiency / CNN design
Top-Down Feedback for Crowd Counting Convolutional Neural Network
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 top-down feedback gating to the bottom-up CNN (ablation) | Crowd-counting CNN · ShanghaiTech Part A | MAE: 147.4 -> 97.5 | Helped |
| fine-tune the bottom-up CNN with count loss instead of adding feedback (ablation) | Bottom-up CNN · ShanghaiTech Part A | MAE: 147.4 -> 116.2 | Helped |
| use top-down feedback rather than count-loss fine-tuning alone (ablation) | Bottom-up CNN · ShanghaiTech Part A | MAE: 116.2 -> 97.5 | Helped |
| add top-down feedback to a single 9 × 9 CNN column (ablation) | Single-column crowd-counting CNN · ShanghaiTech Part A | MAE: 158.5 -> 124.6 | Helped |
| add a 5 × 5 CNN column alongside the 9 × 9 column (ablation) | TDF-CNN · ShanghaiTech Part A | MAE: 124.6 -> 97.5 | Helped |
| use count loss rather than density-map loss to train the top-down network (ablation) | Top-down network · ShanghaiTech Part A | counting performance | Helped |
| add sparse-activation L1 regularization to feedback gates (ablation) | Top-down network · Crowd-counting datasets | counting performance | Helped |
| use top-down feedback on WorldExpo'10 | Crowd-counting CNN · WorldExpo'10 | average MAE: 11.6 -> 11.5 | Helped |
| use TDF-CNN instead of Hydra2s | Crowd-counting model · UCF CC 50 | MAE: 333.7 -> 354.7 | Hurt |
Closest papers by meaning
- Switching Convolutional Neural Network for Crowd Counting (CVPR 2017 · similarity 0.92)
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · similarity 0.92)
- Divide and Grow: Capturing Huge Diversity in Crowd Images With Incrementally Growing CNN (CVPR 2018 · similarity 0.92)
- Context-Aware Crowd Counting (CVPR 2019 · similarity 0.91)
- Shallow Feature Based Dense Attention Network for Crowd Counting (AAAI 2020 · similarity 0.90)
- Attention Scaling for Crowd Counting (CVPR 2020 · 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.