Vision-Based Fall Detection

• 

  MHI Algorithm

The elderly is a fastest growing age group today. Falls among elderly is one of a major concern. A system to detect a fall early would help decreasing the following damage on elderly.

This project utilize GluonCV YOLOv3 Object Detector model to detect the person and uses the Fall Detection model to identify fall and not-fall situation.

The Fall Detection model for this project based on the paper Real-time Vision-based Fall Detection with Motion History Images and Convolutional Neural Networks by T. HARALDSSON. We use Motion History Images (MHIs) algorithm to generate inputs for the model.

The model use Convolutional Network with pretrained MobileNetV2 with PyTorch framework.

The Fall Detecton Dataset FDDis used for training the model

Built With

• pytorch
• sagemaker