rock-paper-scissors ai classifier

This project presents a machine learning solution to the classic game of Rock-Paper-Scissors using image classification. The core of this solution is a Convolutional Neural Network (CNN) designed to recognize hand gestures corresponding to Rock, Paper, and Scissors from static images. The images are preprocessed to a uniform size of 128×128 pixels and labeled accordingly with ‘rock’ as 0, ‘scissors’ as 1, and ‘paper’ as 2.

The CNN model includes multiple layers: two convolutional layers with ReLU activation to capture textural features from the images, followed by max-pooling layers to reduce dimensionality. This is complemented by a flattening step and two dense layers, the last of which uses a softmax activation to classify the images into one of the three categories.

The dataset, comprising labeled images of hand gestures, is split into 80% for training and 20% for testing. After training over 10 epochs, the model achieved an impressive accuracy, demonstrating its ability to generalize well on unseen data.

An additional feature of this project is the simulation of a Rock-Paper-Scissors game where an AI agent plays 100 rounds against predetermined moves, showing a solid win rate based on its predictive capabilities.

The model’s robustness was further tested with random images from the internet, showcasing its potential in real-world scenarios where it could be integrated into interactive games or educational tools.