Research map / Video / Video segmentation and tracking
Spatio-temporal predictive modeling
Models spatiotemporal sequences such as traffic using factorized, Mamba and attention-based architectures.
91 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Video segmentation and tracking, keyword cluster “detection · anomaly · supervised”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
22 papers in 2023–24 and 35 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 5.0% to 6.5%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Predicting the Next Location: A Recurrent Model with Spatial and Temporal Contexts (AAAI 2016 · 993 citations)
- 2017 · Predicting Soccer Highlights from Spatio-Temporal Match Event Streams (AAAI 2017 · 49 citations)
- 2018 · Deep Multi-View Spatial-Temporal Network for Taxi Demand Prediction (AAAI 2018 · 1,104 citations)
- 2019 · Where to Go Next: A Spatio-Temporal Gated Network for Next POI Recommendation (AAAI 2019 · 358 citations)
- 2020 · Self-Attention ConvLSTM for Spatiotemporal Prediction (AAAI 2020 · 384 citations)
- 2021 · GSNet: Learning Spatial-Temporal Correlations from Geographical and Semantic Aspects for Traffic Accident Risk Forecasting (AAAI 2021 · 103 citations)
- 2022 · Decoupling and Recoupling Spatiotemporal Representation for RGB-D-Based Motion Recognition (CVPR 2022 · 50 citations)
- 2023 · Temporal Attention Unit: Towards Efficient Spatiotemporal Predictive Learning (CVPR 2023 · 209 citations)
- 2024 · Hawkes-Enhanced Spatial-Temporal Hypergraph Contrastive Learning Based on Criminal Correlations (AAAI 2024 · 18 citations)
- 2025 · Decomposed Spatio-Temporal Mamba for Long-Term Traffic Prediction (AAAI 2025 · 24 citations)
- 2026 · Spatio-Temporal Context Learning with Temporal Difference Convolution for Moving Infrared Small Target Detection (AAAI 2026 · 8 citations)
Earliest papers here
- Predicting the Next Location: A Recurrent Model with Spatial and Temporal Contexts (AAAI 2016 · 993 citations)
- Spatio-Temporal Hilbert Maps for Continuous Occupancy Representation in Dynamic Environments (NeurIPS 2016 · 17 citations)
- Joint Unsupervised Deformable Spatio-Temporal Alignment of Sequences (CVPR 2016 · 4 citations)
What has been tried
70 papers here report 725 measured changes (a component added, removed or swapped, compared with a baseline): helped 76%, no clear effect 6%, hurt 15%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add temporal decomposition | PEMS08 · MSE | 0.139 -> 0.119 | Helped | Decomposed Spatio-Temporal Mamba for Long-Term Traffic Prediction (AAAI 2025 · 24 citations) |
| Add Hawkes temporal enhancement | NYC and CHI crime prediction · prediction performance | Helped | Hawkes-Enhanced Spatial-Temporal Hypergraph Contrastive Learning Based on Criminal Correlations (AAAI 2024 · 18 citations) | |
| Add temporal tokens | PEMS08 · MAE | 13.81 -> 13.31 | Helped | STD-PLM: Understanding Both Spatial and Temporal Properties of Spatial-Temporal Data with PLM (AAAI 2025 · 17 citations) |
| Removing the hypergraph and feeding raw user sequences directly (STI-HRL') | New York and Tokyo · unspecified (shown in figure) | Helped | Spatial-Temporal Interplay in Human Mobility: A Hierarchical Reinforcement Learning Approach with Hypergraph Representation (AAAI 2024 · 14 citations) | |
| add aligned spatial fixation features instead of misaligned fixation features | Saliency4ASD, subject-wise · accuracy at 0.5 threshold | 0.7433 -> 0.9000 | Helped | SpFormer: Spatio-Temporal Modeling for Scanpaths with Transformer (AAAI 2024 · 13 citations) |
| Add user information to the POI-only model | NYC · Acc@1 | 0.2236 -> 0.2486 | Helped | Integrating Personalized Spatio-Temporal Clustering for Next POI Recommendation (AAAI 2025 · 11 citations) |
| Add temporal augmentation | NYC and TKY (400 users) · model performance | Helped | Scalable Trajectory-User Linking with Dual-Stream Representation Networks (AAAI 2025 · 9 citations) | |
| add self-supervised learning | PeMS08 · MAE | 13.70 -> 12.05 | Helped | SSL-STMFormer Self-Supervised Learning Spatio-Temporal Entanglement Transformer for Traffic Flow Prediction (AAAI 2025 · 8 citations) |
| replace fixed windows with random window offsets | IJRR · MSE | 0.048 -> 0.039 | Helped | EventMamba: Enhancing Spatio-Temporal Locality with State Space Models for Event-Based Video Reconstruction (AAAI 2025 · 6 citations) |
| replace the base positional encoding with FCPE | Medications and Providers · accuracy | 29.3 -> 33.5; 40.2 -> 43.9 | Helped | XTSFormer: Cross-Temporal-Scale Transformer for Irregular-Time Event Prediction in Clinical Applications (AAAI 2025 · 4 citations) |
| replace 1D motion VQVAE with 2D joint VQVAE | HumanML3D · FID | 0.108 -> 0.194 | Hurt | MoGenTS: Motion Generation based on Spatial-Temporal Joint Modeling (NeurIPS 2024 · 4 citations) |
| Add temporal attention | KTH Action · PSNR | 27.7 (-0.5) -> 28.2 | Helped | UniSTD: Towards Unified Spatio-Temporal Learning across Diverse Disciplines (CVPR 2025 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 18 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- PEMS04 · 3 of 18 papers
- PEMS07 · 3 of 18 papers
- PEMS08 · 3 of 18 papers
- PEMS03 · 2 of 18 papers
- TaxiBJ · 2 of 18 papers
Baselines
- iTransformer · 4 of 18 papers
- PatchTST · 2 of 18 papers
- FEDformer · 2 of 18 papers
- Autoformer · 2 of 18 papers
- GWNET · 2 of 18 papers
- MTGNN · 2 of 18 papers
- PDFormer · 2 of 18 papers
- AGCRN · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 57 papers; from OpenAlex and the papers' affiliations).
- University of Electronic Science and Technology of China · 4 papers
- Hong Kong University of Science and Technology · 4 papers
- University of Science and Technology of China · 3 papers
- Chinese Academy of Sciences · 3 papers
- Aalborg University · 2 papers
- Nanjing University · 2 papers
- Southwestern University of Finance and Economics · 2 papers
- Kash Institute of Electronics and Information Industry · 2 papers
- University of Southern California · 2 papers
- Beijing Institute of Technology · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Self-supervised video representation learning (167 papers)
- Video restoration with temporal priors (159 papers)
- Sign language and physiological video analysis (99 papers)
- Weakly supervised video anomaly detection (99 papers)
- Deepfake video detection (47 papers)
Elsewhere in the topic
- Layered video decomposition and dynamics (140 papers)
- Stochastic video prediction (125 papers)
- Physics-inspired motion prediction (113 papers)
- Matching-based video object segmentation (86 papers)
- Learned optical flow estimation (73 papers)
- Learned video compression (68 papers)
- Mask propagation for unsupervised video segmentation (67 papers)
- Video semantic and panoptic segmentation (67 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).