Research map / Video / Video segmentation and tracking
MAU: A Motion-Aware Unit for Video Prediction and Beyond
Zheng Chang, Xinfeng Zhang, Shanshe Wang, Siwei Ma, Yan Ye, Xinguang Xiang, Wen Gao
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 information recalling scheme between encoder and decoder (ablation) | MAU · Moving MNIST | MSE/frame: 29.5 -> 27.6 | Improved |
| add information recalling scheme (ablation) | MAU · Caltech Pedestrian | MSE (10^-3): 1.34 -> 1.24 | Improved |
| add information recalling scheme (ablation) | MAU · Caltech Pedestrian | FVD: 1269.9 -> 1204.0 | Improved |
| add information recalling scheme (ablation) | MAU · TownCentreXVID t=5 | PSNR: 30.61 -> 31.87 | Improved |
| increase temporal receptive field tau from 1 to 5 (ablation) | MAU · Moving MNIST | MSE/frame: 33.4 -> 29.5 | Improved |
| increase temporal receptive field tau from 5 to 10 (ablation) | MAU · Moving MNIST | MSE/frame: 29.5 -> 29.3 | No clear effect |
| increase temporal receptive field tau from 1 to 5 (ablation) | MAU · Moving MNIST | inference time: 14.90s -> 17.36s | Hurt |
| replace MotionGRU with MAU (same encoder/decoder) (ablation) | MAU · Moving MNIST | MSE: 34.3 -> 29.5 | Improved |
| replace E3D-LSTM with MAU (ablation) | MAU · Moving MNIST | inference time: 57.21s -> 17.34s | Improved |
| use MAU as predictive unit instead of MotionGRU | MAU · Something-Something V2 front 50% | top-1 accuracy: 10.22 -> 10.60 | 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 1 of 1 shared result
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
- STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond (TPAMI 2025 · similarity 0.88)
- Memory-augmented Dense Predictive Coding for Video Representation Learning (ECCV 2020 · similarity 0.86)
- Video Prediction Recalling Long-Term Motion Context via Memory Alignment Learning (CVPR 2021 · similarity 0.85)
- Memory-Augmented Temporal Dynamic Learning for Action Recognition (AAAI 2019 · similarity 0.85)
- DynamoNet: Dynamic Action and Motion Network (ICCV 2019 · similarity 0.85)
- Temporal Attention Unit: Towards Efficient Spatiotemporal Predictive Learning (CVPR 2023 · 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.
