Rock-paper-scissors game using real-time object detection

Lim, Xiao Yun (2025) Rock-paper-scissors game using real-time object detection. Final Year Project, UTAR.

Preview

PDF Download (2671Kb) | Preview

Abstract

This project introduces an innovative Rock-Paper-Scissors (RPS) game that integrates real-time hand gesture recognition within a Flutter-based mobile application, leveraging advanced machine learning techniques. Utilizing MobileNetV2, a lightweight convolutional neural network, the system reliably classifies rock, paper, and scissors gestures from live camera feeds. Developed through an evolutionary prototyping methodology, the project iteratively refined a TensorFlow Lite-deployed model and a user-friendly interface featuring tutorial screens, game history tracking, and celebratory animations. OpenCV ensured robust dataset preprocessing, enabling high-quality training data, while Flutter facilitated seamless cross-platform performance. Extensive testing confirmed the system’s effectiveness across diverse lighting conditions and device specifications, achieving consistent gesture detection and rapid UI responsiveness. By addressing challenges such as gesture variability and real-time processing latency through model optimization and efficient camera handling, the project delivers an immersive gaming experience without physical controllers. This work advances interactive gaming by demonstrating the feasibility of deploying sophisticated machine learning models on resource-constrained mobile devices. The framework offers potential for applications in educational tools and assistive technologies, contributing to further developments in computer vision and human-computer interaction. Area of Study (Minimum 1 and Maximum 2): Deep Learning, Mobile Computing Keywords (Minimum 5 and Maximum 10): Gesture Recognition, Machine Learning, CNN, Mobile Application, Real-time Processing

Actions (login required)