Research map / Detection and segmentation / Tracking
Fusing Crowd Density Maps and Visual Object Trackers for People Tracking in Crowd Scenes
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 a sparsity constraint to KCF response maps | KCF · UCSD | P@10: 0.4235 -> 0.4356 | Helped |
| Add a sparsity constraint to KCF response maps | KCF · PETS2009 | P@10: 0.2251 -> 0.2447 | Helped |
| Fuse tracker response maps with crowd density maps using FusionCNN-v2 | S-KCF · UCSD | P@10: 0.4356 -> 0.5999 | Helped |
| Fuse tracker response maps with crowd density maps using FusionCNN-v2 | S-KCF · PETS2009 | P@10: 0.2447 -> 0.3326 | Helped |
| Fine-tune the batch-trained fusion model using the real frame-by-frame tracking process (ablation) | FusionCNN · UCSD | P@10: 0.5772 -> 0.5999 | Helped |
| Fine-tune the batch-trained fusion model using the real frame-by-frame tracking process (ablation) | FusionCNN · PETS2009 | P@10: 0.2813 -> 0.3326 | Helped |
| Add color cues to HOG features | S-KCF · UCSD | P@10: 0.4356 -> 0.5259 | Helped |
| Add color cues to HOG features | S-KCF · PETS2009 | P@10: 0.2447 -> 0.2529 | Helped |
| Use a single fusion model across crowd levels rather than separate per-level models (ablation) | S-KCF · PETS2009 | P@10: 0.3326 -> 0.3245 | Hurt |
| Train the fusion model on UCSD and transfer it to PETS2009 (ablation) | S-KCF · PETS2009 | P@10: 0.2447 -> 0.1163 | Hurt |
| Use the three-layer fusion CNN instead of a two- or four-layer version (ablation) | FusionCNN with S-KCF · UCSD | tracking performance | Helped |
| Include image patch, crowd density map, and tracker response map as all three fusion inputs (ablation) | FusionCNN with S-KCF · UCSD | tracking performance | Helped |
Closest papers by meaning
- Tracking Pedestrian Heads in Dense Crowd (CVPR 2021 · similarity 0.87)
- Divide and Grow: Capturing Huge Diversity in Crowd Images With Incrementally Growing CNN (CVPR 2018 · similarity 0.87)
- Top-Down Feedback for Crowd Counting Convolutional Neural Network (AAAI 2018 · similarity 0.86)
- Multi-view Crowd Tracking Transformer with View-Ground Interactions Under Large Real-World Scenes (CVPR 2026 · similarity 0.86)
- 3D Crowd Counting via Multi-View Fusion with 3D Gaussian Kernels (AAAI 2020 · similarity 0.86)
- CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · similarity 0.84)
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.