A Method for Clustering and Classifying Hand shapes from Auslan

How to use

Step 1:

Use the get_signs.py script to create your data set of Auslan signs.

Step 2:

Then use the create_data.py script to run the MediaPipe framework over your collected data to create a dataset for training a fully connected network.

Step 3:

Next, use the make_hand_model.py script to train a full connected network. Be careful to examine the directory references.

Step 4:

Finally, Use posenet_video.py script to classify your handshapes from data has been collected. Make sure this script references your trained hand model

Optional

For accuracy and performance metrics, use the classification_report.py script to see your model's accuracy

Scripts that end in "clustering.py" and "crop_hand.py" are used for clustering from a video source, where the number of hand shapes are unknown.