Research map / Video / Video segmentation and tracking
Temporal Attention Unit: Towards Efficient Spatiotemporal Predictive Learning
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 |
|---|---|---|---|
| Use TAU instead of recurrent or convolutional baseline methods | spatiotemporal prediction · Moving MNIST | MSE: 22.3 -> 19.8 | Helped |
| Use TAU instead of recurrent or convolutional baseline methods | spatiotemporal prediction · TaxiBJ | MSE × 100: 41.4 -> 34.4 | Helped |
| Use TAU instead of recurrent or convolutional baseline methods | spatiotemporal prediction trained on KITTI · Caltech Pedestrian | SSIM: 0.940 -> 0.946 | Helped |
| Use TAU instead of recurrent or convolutional baseline methods | spatiotemporal prediction · KTH (10→40) | PSNR: 32.93 -> 33.01 | Helped |
| Replace TAU with vanilla 3×3 convolutional blocks (ablation) | TAU · Moving MNIST | MSE: 19.8 -> 58.9 | Hurt |
| Replace TAU with vanilla 3×3 convolutional blocks (ablation) | TAU · TaxiBJ | MSE × 100: 34.4 -> 43.5 | Hurt |
| Add static attention (SA) (ablation) | TAU · Moving MNIST | MSE: 23.2 -> 19.8 | Helped |
| Add static attention (SA) (ablation) | TAU · TaxiBJ | MSE × 100: 40.8 -> 34.4 | Helped |
| Add dynamic attention (DA) (ablation) | TAU · Moving MNIST | MSE: 22.4 -> 19.8 | Helped |
| Add dynamic attention (DA) (ablation) | TAU · TaxiBJ | MSE × 100: 38.4 -> 34.4 | Helped |
| Add differential divergence regularization (DDR) (ablation) | TAU · Moving MNIST | MSE: 21.1 -> 19.8 | Helped |
| Add differential divergence regularization (DDR) (ablation) | TAU · TaxiBJ | MSE × 100: 37.7 -> 34.4 | Helped |
Closest papers by meaning
- Wavelet-Driven Spatiotemporal Predictive Learning: Bridging Frequency and Time Variations (AAAI 2024 · similarity 0.89)
- PredRNN: Recurrent Neural Networks for Predictive Learning using Spatiotemporal LSTMs (NeurIPS 2017 · similarity 0.88)
- UniSTD: Towards Unified Spatio-Temporal Learning across Diverse Disciplines (CVPR 2025 · similarity 0.85)
- State Space Prompting via Gathering and Spreading Spatio-Temporal Information for Video Understanding (NeurIPS 2025 · similarity 0.85)
- STOP: Integrated Spatial-Temporal Dynamic Prompting for Video Understanding (CVPR 2025 · similarity 0.85)
- STDiff: Spatio-Temporal Diffusion for Continuous Stochastic Video Prediction (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.