This repository houses the machine learning/neural network implementation of the Universal Gestures project. See the Technical Specification for more details.
Related: Universal Gestures Unity Project and Scrum Board.
Install Python 3.12.3 or later.
From the root directory, run the following command to install the required packages:
pip install -r requirements.txt
- Populate
data/
with json data collected from the Unity data collection scene. - Run
process_data.py
to split the dataset into test and train. - Run
model.py
to train the model, or runmodel_two_hands.py
to train a two-handed model. Ensure the number of input features to the model is correct. - Find the outputted weights in
trained_model/
. The model is outputted in both .onnx and .json formats. The onnx model can then be imported into the Universal Gestures Unity project and used in the inference script.