Adaptive 3D Gameplay with Dynamic Generative Gaming using Three.js and Semi-Supervised Learning

This paper introduces a generative gaming system that uses Three.js for real-time 3D visuals combined with semi-supervised learning. The system dynamically generates game content, such as levels and characters, based on both labeled data and real-time player interactions. By leveraging Three.js, the game ensures smooth transitions between AI-driven content creation and high-quality 3D rendering. The goal is to enhance player engagement through a dynamic and adaptive gameplay experience. The project explores how natural language prompts can influence the game's storyline and environment, allowing for a more personalized experience. A generative model is trained on a dataset of player interactions, which enables the creation of game scenarios tailored to individual preferences. This semi-supervised learning framework combines labeled and unlabeled data to understand player behaviour better. Preliminary results suggest that the generative approach leads to a more immersive and personalized gameplay experience, with positive feedback highlighting its effectiveness. The system architecture efficiently manages the synchronization between game generation and 3D rendering, addressing challenges related to data representation and model performance. In conclusion, this project showcases the potential of semi-supervised learning in gaming, improving player engagement and re-playability by creating adaptive, dynamic game worlds.