Links to understand further:
Pre-requisites:
Understanding CNN, RNN, LSTM architecture
Creating Video Dataset
-
Create 5 videos under non-falling section (say Warm-up, Jump, Burpees, Walking, Push-ups, Sitting) and One falling video each of 60 seconds
-
Use the same camera for all video creation and convert them into 640 X 480 Pxl for our project
-
Camera used - iPhone SE 12 MP which gives 1920 X 1080
-
Click the image below to view for falling video
-
For Data set to do this project - Click here
-
Installing necessary Libraries
-
create conda environment for our project
conda create -n falldetection pip scipy pyyaml python=3.6
-
Activate conda environment
conda activate falldetection
-
Install opencv-python and tensorflow 1.14 version
pip install tensorflow==1.14 pip install opencv-python
-
Clone Posenet link from github to local. Run webcam.py file. It must open the webcam and check whether it's identifying the skeletal points
-
--To be Updated--
References
- Understanding CNN architecture
- --To be Updated--