Research map / Video / Video segmentation and tracking
S2F2: Single-Stage Flow Forecasting for Future Multiple Trajectories Prediction
Yu-Wen Chen, Hsuan-Kung Yang, Chu-Chi Chiu, Chun-Yi Lee
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 |
|---|---|---|---|
| replace two-stage STED with single-stage S2F2 | trajectory forecasting · StaticMOT val forecasting (3 future frames) | FDE: 23.761 -> 16.228 | Improved |
| replace two-stage STED with single-stage S2F2 | trajectory forecasting · StaticMOT val forecasting (3 future frames) | latency (ms): 623.480 -> 13.788 | Improved |
| replace constant velocity with S2F2 | trajectory forecasting · StaticMOT val forecasting (3 future frames) | ADE: 14.481 -> 12.275 | Improved |
| add SSIM loss (ablation) | S2F2 · MOT17 val (trained on MOT17) ablation | FDE: 43.021 -> 39.891 | Improved |
| add SSIM loss (ablation) | S2F2 · MOT17 val (trained on MOT17) ablation | MOTA: 67.0 -> 67.7 | Improved |
| add past flow decoder (ablation) | S2F2 · MOT17 val (trained on MOT17) ablation | FDE: 39.891 -> 43.097 | Hurt |
| add past flow decoder with detached past GRU encoder (ablation) | S2F2 · MOT17 val (trained on MOT17) ablation | FDE: 39.891 -> 46.869 | Hurt |
| add forecasting refinement to online association (ablation) | S2F2 · MOT17 | MOTA: 67.6 -> 67.7 | Improved |
| add forecasting refinement to online association (ablation) | S2F2 · MOT17 | FDE: 40.014 -> 39.891 | Improved |
| add forecasting refinement to online association (ablation) | S2F2 · StaticMOT | MOTA: 73.3 -> 73.6 | Improved |
| replace FairMOT association with ByteTrack (ablation) | S2F2 · MOT17 | MOTA: 67.7 -> 66.2 | Hurt |
| replace FairMOT association with ByteTrack (ablation) | S2F2 · StaticMOT | MOTA: 73.6 -> 73.4 | Hurt |
Closest papers by meaning
- MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation (CVPR 2025 · similarity 0.87)
- FlowMoP: Stochastic Multi-Person Motion Prediction (NeurIPS 2026 · similarity 0.87)
- GigaTraj: Predicting Long-term Trajectories of Hundreds of Pedestrians in Gigapixel Complex Scenes (CVPR 2024 · similarity 0.86)
- MESSENGER: Memory-Enhanced Sequential Scene Flow Estimation via Autoregressive Next-Frame Forecasting (NeurIPS 2026 · similarity 0.85)
- MS-TIP: Imputation Aware Pedestrian Trajectory Prediction (ICML 2024 · similarity 0.85)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.85)
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.
