CSI2PointCloud: Spatio-Temporal 3D Point Clouds from WiFi-CSI Data via Transformer Networks

Описание: This project aims to estimate 3D point clouds of indoor environments using non-contact sensing via WiFi Channel State Information (CSI) data obtained through Multiple-Input Multiple-Output (MIMO) configurations. By leveraging a transformer-based architecture, the system processes temporal CSI data, including both amplitude and phase information, from multiple transmitters and receivers to generate accurate and detailed 3D reconstructions of indoor spaces.

The approach effectively captures the spatial and temporal dynamics of the environment, enabling modeling of human presence and movement within various indoor settings. The system is evaluated on the MM-Fi dataset [1], which comprises comprehensive CSI measurements reflecting human activities across diverse indoor environments.

This repository contains the implementation code for the paper titled “Spatio-Temporal 3D Point Clouds from WiFi-CSI Data via Transformer Networks” available on arXiv. You can access the full details and experimental results in the published work.

Paper: https://arxiv.org/pdf/2410.16303
Repository: https://github.com/Arritmic/csi2point...
Citation: Määttä, Tuomas, Sasan Sharifipour, Miguel Bordallo López, and Constantino Alvarez Casado. "Spatio-Temporal 3D Point Clouds from WiFi-CSI Data via Transformer Networks." arXiv preprint arXiv:2410.16303 (2024).

Music: Arritmic (Tino Álvarez)