Static hand gesture recognition are the leading problems in many industries like gaming, virtual reality, robot control, interactive sign language, gesture recognition e.t.c. Here what static means where hand remains still in a certain Pose like closing Fist, point with 1 finger e.t.c. In this project our task is to classify the gesture based on the axial dataset which is generated by the Vicon motion capture camera system which has 36 dimensional data. Existing methodology of classification has the best balanced error rate close to 0.158. Our aim is to decrease it further, by introducing some new machine learning techniques. We provide the best ensemble technique along with deep neural networks which are capable of doing so.
This project is python flask based API exposed to the outside world for use.
requirement
- python enterprator version 3.7+
how to run
- clone th repository
- go to the root location i.e. posture_prediction_app
- Create virtual environment by command "python -m venv .\venv" (.\venv is the current location for virtual environment)
- pip install --upgrade pip (better to upgrade pip to the letest version)
- pip install -r requirement.txt
- .\main.py (start the server)
Test the running API http://localhost:8080/api/v1/predict-posture should return { msg: "Test Endpoint for predict_posture" }
API is running.you can test rest of the end points specify in src/v1/end_points for predicting class or probability