SilverAvocado/Silver-Multimodal-Dataset

该数据集旨在支持从音频和视频源中检测日常活动、暴力和跌倒场景的机器学习模型的开发。预处理流程包括音频特征提取、人体关键点检测和相对位置编码，生成统一的表示用于训练和推理。数据存储为`.npy`文件，分为三类：日常活动、暴力行为和跌倒事件。数据集结合了MFCC音频特征和MediaPipe关键点，确保模型能够准确检测和区分定义的活动类别。

The dataset is designed to support the development of machine learning models for detecting daily activities, violence, and fall down scenarios from combined audio and video sources. The dataset generates a unified representation for training and inference through audio feature extraction, human keypoint detection, and relative positional encoding. It contains three classes: daily activities, violent behaviors, and fall down events. The data is stored in .npy files, each containing concatenated audio and video feature representations for a fixed sequence of frames. The preprocessing includes multi-step audio and video processing, and the final features are concatenated and saved in .npy format. The dataset is suitable for safety systems, healthcare monitoring, and smart home applications.