Skip to content

elsakoh/cs3244-g26

Repository files navigation

Fall-Net

Classify videos of human activity as fall or not-fall events. CS3244 AY20/21 Sem 2 Project.

This repository uses python 3.x and is tested on python 3.6

Datasets

Model has been trained on these public datasets: URFD and MCFD.

Training and testing datasets used

  • SISFD
  • NTU-RGBD, only the FALL and ADL (Activities of Daily living) datasets

Project structure

Preprocessing

Video and Image Preprocessing files are found in dataset_preprocessing/. preprocessing is split into two parts:

  1. formatting videos into frames preprocess_XX.py
  2. generating optical flow images generate_OF_XX.py

The preprocessed images are used as inputs to train the model.

Model

Refer to file temporalnet.py. Each model works specifically on each of the fall datasets.

Refer to paper cited below for the model's architecture.

Running the model

  1. Install dependencies pip3 install -r requirements.txt
  2. python temporalnet.py

References:

Fall-Detection-with-CNNs-and-Optical-Flow

This repository contains code from the paper:

Núñez-Marcos, A., Azkune, G., & Arganda-Carreras, I. (2017).
"Vision-Based Fall Detection with Convolutional Neural Networks"
Wireless Communications and Mobile Computing, 2017.

Contributors:

Abhinav Ramnath , Carel Chay Jia Ming, Koh Hui Hui Elizabeth Fang Pin Sern, Goh Jia Jun, Jordan Rahul Sabanayagam

About

ML Singapore! Group project

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages