Falls are one of the leading causes of injury in older persons. With around 27,000 seniors succumbing annually due to falls, there is a glaring need to confront this issue. We created an adaption of the Temporal Shift Module developed by the MIT HAN Lab to detect human falls, assessing the feasibility for a home camera system to detect falls as they occur.
- Python 3.5 or greater
- Jetson TX2
- Docker
- Build the docker image and run your container
docker built -t tsm_test1 .
docker run --privileged -v $(pwd):/falling -v /dev/bus/usb:/dev/bus/usb -v /tmp:/tmp -e QT_X11_NO_MITSHM=1 -e DISPLAY=$DISPLAY -it tsm_test1 bash
- The demo script can be run either using the camera feed or using a local video
python3 main.py --video ./test_video/test_fall.avi
To use the camera, please exclude the --video option