Rock, Paper, Scissors - Using CNN

Play Rock, Paper, Scissors game with computer using your camera.

Dependencies

1. Keras - Tensorflow
2. OpenCV
3. numpy

The model was trained to detect number of fingers presented(zero, one, two, three, four, five) using Convolutional Neural Network that was built using Keras. This model was used to predict the user input and play the game. The training had an accuracy of 80% which was fairly good. The number of samples that I made was less and also not very dynamic.

TODOs

1. Make the model more robust to background. Currently performs well only in white background.
2. Increase the training dataset.
3. Upload the training dataset.