An ML model which can detect activities of daily life like Sit,Stand,Bend,Trip,Fall. Model detects human fall position take snaps at the incident and sends it to the nodejs application for storing. The model is trained using https://teachablemachine.withgoogle.com/train/pose. Do train on a different environment the model can fail on a different data distribution.
Install Nodejs
git clone https://github.com/TheG00dB0y/humanadl_fall_detection
cd ./humanadl_fall_detection
//installing node modules
npm install
//Running node application at 127.0.0.1 port 80
node index
//Open a browser and send the GET request to 127.0.0.1
Change ALERT_THRESH value to change the frequency in which snaps are taken on detection of an alert
Currently alert is defined as "Stand" for an easy demo, do change the ALERT_CLASS to required class name (Available classes: Sit,Stand,Bend,Trip,Fall)
Make sure to enter '127.0.0.1 ' in the browser address bar due to CORS policy error. (AJAX POST request is send to that address/server)
Chrome is not setting the img src as empty works fine with firefox . The model do face issue the lack of data and environment changes. Do retrain further to improve the accuracy.