Research map / Video / Video segmentation and tracking
MIMO Is All You Need:A Strong Multi-in-Multi-Out Baseline for Video Prediction
Shuliang Ning, Mengcheng Lan, Yanran Li, Chaofeng Chen, Qian Chen, Xunlai Chen, Xiaoguang Han, Shuguang Cui
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 |
|---|---|---|---|
| Use MIMO-VP instead of CrevNet | Video prediction on Moving MNIST 10-10 · Moving MNIST 10-10 | MSE: 22.3 -> 17.7 | Improved |
| Use MIMO-VP instead of PhyDNet | Video prediction on Moving MNIST 10-30 · Moving MNIST 10-30 | MSE: 58.7 -> 31.6 | Improved |
| Use MIMO-VP instead of PhyDNet | Video prediction on Human3.6M · Human3.6M | MAE/100: 16.2 -> 10.4 | Improved |
| Use MIMO-VP instead of LMC-Memory | Video prediction on KITTI · KITTI | MAE/100: 44.6 -> 40.9 | Improved |
| Use MIMO-VP instead of PhyDNet | Video prediction on Weather · Weather | MAE: 280.1 -> 235.6 | Improved |
| Add 2D convolution multi-head attention (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MSE: 32.8 -> 17.7 | Improved |
| Add 2D convolution multi-head attention (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MAE: 84.4 -> 51.6 | Improved |
| Add the local spatio-temporal block (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MSE: 21.2 -> 17.7 | Improved |
| Add the local spatio-temporal block (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MAE: 59.2 -> 51.6 | Improved |
| Use parallel multi-output prediction rather than recursive single-output prediction (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MSE: 22.8 -> 17.7; 18.0% | Improved |
| Use parallel multi-output prediction rather than recursive single-output prediction (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MAE: 62.8 -> 51.6; 14.0% | Improved |
| Keep decoder self-attention to model dependencies among future frames (ablation) | MIMO-VP on Moving MNIST · Moving MNIST | MSE: 21.6 -> 17.7 | 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.
- SimVP — SimVP: Simpler Yet Better Video Prediction (CVPR 2022) · 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
- Flexible Spatio-Temporal Networks for Video Prediction (CVPR 2017 · similarity 0.83)
- MOSO: Decomposing MOtion, Scene and Object for Video Prediction (CVPR 2023 · similarity 0.82)
- STDiff: Spatio-Temporal Diffusion for Continuous Stochastic Video Prediction (AAAI 2024 · similarity 0.81)
- Motion Graph Unleashed: A Novel Approach to Video Prediction (NeurIPS 2024 · similarity 0.81)
- Video Prediction by Modeling Videos as Continuous Multi-Dimensional Processes (CVPR 2024 · similarity 0.80)
- MIMO: Controllable Character Video Synthesis with Spatial Decomposed Modeling (CVPR 2025 · similarity 0.80)
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.
