All Examples are in the directory and demonstrated in the python notebook!
You need to have a camera connected to your machine and functioning correctly
python -m venv myenv
source myenv/bin/activate
-
I used a pip mirror in China since I am travelling, but you don't have to
Simply change the pip install line to:
!pip install tensorflow==2.8.1 tensorflow-gpu==2.8.1 opencv-python mediapipe scikit-learn matplotlib
-
The gpu option is disabled since I am away from my computer that can run models with gpu. I encourage you to try to turn it on since it is alot faster to train.
-
The gestures I tried to classify are hello, no, and iloveyou. However, you can modify the # of class in:
# Actions that we try to detect
actions = np.array(['hello', 'no', 'iloveyou'])
-
You can adjust the # of training samples here:
# Thirty videos worth of data
no_sequences = 30
Remember to clear the memory and free up the camera once you stop the training process by:
cap.release()
cv2.destroyAllWindows()