Multi-Model Deep Learning for Rock-Paper-Scissors Gesture Recognition

This project focuses on building a deep learning-based hand gesture recognition system that classifies images of the Rock, Paper, and Scissors gestures using PyTorch. The objective is to implement and compare the performance of three different neural network architectures: an Artificial Neural Network (ANN), a Convolutional Neural Network (CNN), and a simplified Residual Network (ResNet).

The dataset used consists of CGI-generated images of hand gestures with variations in background, lighting, and angle to mimic real-world conditions. The workflow includes dataset preparation, image preprocessing, and data augmentation techniques to enhance the model's generalization capabilities. Each model is trained and evaluated using accuracy metrics and confusion matrices. TensorBoard is integrated for visualizing training and validation performance.

The ANN model serves as a baseline with fully connected layers, while the CNN model leverages spatial hierarchies for feature extraction. The ResNet model introduces skip connections to address the vanishing gradient problem and enables deeper architectures.

This project highlights the comparative strengths of each model type in the context of image classification and demonstrates the use of PyTorch for end-to-end model development, training, and evaluation. The final results are visualized and interpreted to determine the most effective architecture for gesture recognition tasks.