Research map / Video / Video segmentation and tracking
Laying the Foundations of Deep Long-Term Crowd Flow Prediction
Samuel S. Sohn, Honglu Zhou, Seonghyeon Moon, Sejong Yoon, Vladimir Pavlovic, Mubbasir Kapadia
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 |
|---|---|---|---|
| train on all compression rates and density levels instead of a single rate (CR 1.0) (ablation) | CAGE + SegNet · Proxy flow, test CR 2.0 | MAE: 0.0728 -> 0.0317 | Improved |
| train on all compression rates instead of CR 1.0 only (ablation) | CAGE + SegNet · Proxy flow, test CR 1.5 | MAE: 0.0577 -> 0.0273 | Improved |
| train only at CR 2 instead of CR 1.0 (ablation) | CAGE + SegNet · Proxy flow, test CR 1.0 | MAE: 0.0164 -> 0.1011 | Hurt |
| train at CR 1.25 or CR 1.75 instead of CR 1.0 only (ablation) | CAGE + SegNet · Proxy flow across test CRs | MAE | Improved |
| use proxy crowd flow instead of Social Force simulation as training data | CAGE + SegNet · Stanford Crowd Flow | MAE | No clear effect |
| add 50% multi-goal non-uniform SF training data (ablation) | PanLSF · MGNU SF scenarios | MAE | Improved |
| generate proxy flow instead of SF simulation | Proxy crowd flow · dataset generation time | time: 167 hours -> little over 2 hours | Improved |
| train only on axis-aligned environments and test on non-axis-aligned real floorplans | CAGE + SegNet · NARF dataset | MAE | No clear effect |
| use CAGE instead of nearest-neighbor, area, bilinear, bicubic, Lanczos resampling | CAGE · environment compression | artifacts | Improved |
Closest papers by meaning
- Generic and Dynamic Graph Representation Learning for Crowd Flow Modeling (AAAI 2023 · similarity 0.88)
- DeepSTN+: Context-Aware Spatial-Temporal Neural Network for Crowd Flow Prediction in Metropolis (AAAI 2019 · similarity 0.87)
- Explainable Origin-Destination Crowd Flow Interpolation via Variational Multi-Modal Recurrent Graph Auto-Encoder (AAAI 2024 · similarity 0.86)
- Continuous Locomotive Crowd Behavior Generation (CVPR 2025 · similarity 0.85)
- Learning Extremely High Density Crowds as Active Matters (CVPR 2025 · similarity 0.84)
- Social Physics Informed Diffusion Model for Crowd Simulation (AAAI 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.
