# VMRNN: Integrating Vision Mamba and LSTM for Efficient and Accurate Spatiotemporal Forecasting

**Authors:** Yujin Tang, Peijie Dong, Zhenheng Tang, Xiaowen Chu, Junwei Liang  
**arXiv:** [2403.16536](https://arxiv.org/abs/2403.16536)  
**Date:** 2024-03-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2403.16536v2/

## Key points

- Paper title: "VMRNN: Integrating Vision Mamba and LSTM for Efficient and Accurate Spatiotemporal Forecasting"
- Authors: Yujin Tang, Peijie Dong, Zhenheng Tang, Xiaowen Chu, Junwei Liang
- Introduction of a novel architecture combining Vision Mamba blocks with LSTM for spatiotemporal forecasting
- VMRNN cell designed to outperform established vision models in efficiency and accuracy
- Utilization of recent advancements in Mamba-based architectures
- Competitive results demonstrated across various tasks with a smaller model size
- Contribution to advancing spatiotemporal forecasting techniques
- Availability of code on GitHub for further exploration

## Summary

In their paper titled "VMRNN: Integrating Vision Mamba and LSTM for Efficient and Accurate Spatiotemporal Forecasting," authors Yujin Tang, Peijie Dong, Zhenheng Tang, Xiaowen Chu, and Junwei Liang introduce a novel architecture that combines the strengths of Vision Mamba blocks with LSTM to tackle spatiotemporal prediction tasks effectively. The proposed VMRNN cell aims to surpass established vision models in both efficiency and accuracy by leveraging recent advancements in Mamba-based architectures. Through extensive evaluations, the authors demonstrate competitive results across a variety of tasks while maintaining a smaller model size. This innovative research contributes to advancing spatiotemporal forecasting techniques and provides an opportunity for further exploration through the availability of their code on GitHub.