The WiFi-Based Human Activity Recognition Under Everyday Life Changes Dataset (WiFi-Wild)

This repository contains the files required to support WiFi-based Human Activity Recognition (HAR) Included Files WiFi-Wild-HAR.zip (2.40 GB)The primary dataset contains raw CSI recordings collected in a real-world environment. This file forms the core resource for training and evaluating HAR model. ESP32-CSI-Tool-master.zip (79.28 KB)The official ESP32 CSI tool used for data acquisition. It includes the necessary scripts and firmware for capturing CSI measurements from ESP32 devices. Source_code.ipynb (9.77 KB)A Jupyter Notebook providing example code for data preprocessing, feature extraction, and model experimentation. This serves as a starting point for reproducing the experiments. Fields_Description.pdf (71.03 KB)Documentation describing the dataset structure, variable names, and field meanings to help users correctly interpret the data. Baseline_Model.kerasA pre-trained deep learning baseline model saved in the native Keras (.keras) format. This model is trained on the baseline layout to be used in testing on different setups available in the WiFi-Wild dataset folder.本仓库包含支持基于WiFi的人体活动识别（Human Activity Recognition, HAR）所需的全部文件。 包含文件 WiFi-Wild-HAR.zip（2.40 GB）：核心数据集，包含真实场景中采集的原始信道状态信息（Channel State Information, CSI）录制数据，是训练与评估HAR模型的核心资源。 ESP32-CSI-Tool-master.zip（79.28 KB）：用于数据采集的官方ESP32 CSI工具，内置从ESP32设备捕获CSI测量值所需的脚本与固件。 Source_code.ipynb（9.77 KB）：一份Jupyter Notebook，提供了数据预处理、特征提取与模型实验的示例代码，可作为复现相关实验的起点。 Fields_Description.pdf（71.03 KB）：数据集说明文档，用于阐释数据集结构、变量名称与字段含义，帮助用户正确解读数据内容。 Baseline_Model.keras：一份以Keras原生（.keras）格式保存的预训练深度学习基线模型，该模型基于基线布局训练完成，可用于在WiFi-Wild数据集文件夹内的不同配置场景中开展测试。