Rock-Paper-Scissors Recognition

Application of Computer Vision to the classic game Rock-Paper-Scissors. This repository includes dataset analysis and modeling part of the task.

The model will be deployed as a JavaScript application on my website.

Dataset: https://www.kaggle.com/frtgnn/rock-paper-scissor

Installation

poetry install
cd data
kaggle datasets download --unzip frtgnn/rock-paper-scissor

Modeling

The best scores I was able to achieve with the following configs:

• Fully Freezed MobileNetV2 + RMSProp + L2Regularization(0.01) - ?
• Fully Freezed MobileNetV2 + AdamW(weight_decay: 0.01) - ?

References

• https://keras.io/guides/transfer_learning/
• https://www.tensorflow.org/tutorials/images/transfer_learning
• https://ruder.io/transfer-learning/
• http://www.laurencemoroney.com/rock-paper-scissors-dataset/