UR-FALL-GS: UR Fall Detection using GSTCAN and Spatiotemporal Analysis

This project implements a fall detection system using the UR Fall Dataset and a Graph-based Spatiotemporal Convolutional Attention Network (GSTCAN). It processes depth images to detect falls in real-time scenarios.

Features

Automatic download and preprocessing of the UR Fall Dataset

Implementation of GSTCAN model for fall detection

Training pipeline with checkpointing and resume capability

Real-time fall detection inference