Research map / Video / Video segmentation and tracking
USTEP: Spatio-Temporal Predictive Learning Under a Unified View
Cheng Tan, Jue Wang, Zhangyang Gao, Siyuan Li, Stan Ziqing Li
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 cross-segment mechanism (ablation) | USTEP w/TAU · Moving MNIST (10->10) | MSE: 24.01 -> 21.84 | Improved |
| use stride dt=5 instead of dt=1 (ablation) | USTEP w/TAU · Moving MNIST (10->10) | MSE: 31.94 -> 21.84 | Improved |
| use stride dt=5 instead of dt=10 (ablation) | USTEP w/TAU · Moving MNIST (10->10) | MSE: 25.13 -> 21.84 | Improved |
| use stride dt=5 instead of dt=2 (ablation) | USTEP w/TAU · Moving MNIST (10->10) | MSE: 24.62 -> 21.84 | Improved |
| use stride dt=5 instead of dt=1 (ablation) | USTEP w/TAU · Moving MNIST (10->10) | training time (s/epoch): 600 -> 93 | Hurt |
| wrap TAU in USTEP | TAU · Moving MNIST (10->10) | MSE: 24.60 -> 21.84 | Improved |
| wrap Uniformer in USTEP | Uniformer · Moving MNIST (10->10) | MSE: 30.38 -> 28.51 | Improved |
| wrap SimVP in USTEP | SimVP · Moving MNIST (10->10) | MSE: 32.15 -> 30.04 | Improved |
| wrap TAU in USTEP | TAU · Moving MNIST (10->10) | Params (M): 44.7 -> 18.9 | Improved |
| wrap TAU in USTEP | TAU · KTH (10->40) | FVD: 1500.63 -> 849.16 | Improved |
| wrap TAU in USTEP | TAU · KTH (10->20) | FLOPs (G): 73.8 -> 107.0 | Hurt |
| wrap TAU in USTEP | TAU · SEVIR (13->12) | CSI: 0.3941 -> 0.4366 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PredRNN++ — PredRNN: Recurrent Neural Networks for Predictive Learning using Spatiotemporal LSTMs (NeurIPS 2017) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- UniSTD: Towards Unified Spatio-Temporal Learning across Diverse Disciplines (CVPR 2025 · similarity 0.86)
- ST-ReP: Learning Predictive Representations Efficiently for Spatial-Temporal Forecasting (AAAI 2025 · similarity 0.86)
- Wavelet-Driven Spatiotemporal Predictive Learning: Bridging Frequency and Time Variations (AAAI 2024 · similarity 0.86)
- STRAP: Spatio-Temporal Pattern Retrieval for Out-of-Distribution Generalization (NeurIPS 2025 · similarity 0.85)
- How Different from the Past? Spatio-Temporal Time Series Forecasting with Self-Supervised Deviation Learning (NeurIPS 2025 · similarity 0.85)
- Diffusion Transformers as Open-World Spatiotemporal Foundation Models (NeurIPS 2025 · 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.
