Highlights

This project aims to create an immersive virtual tourism platform centered on Hong Kong's Big Buddha using 3D Gaussian Splatting technology and an improved Gaussian fitting algorithm. The project involved data collection, 3D modeling, VR environment construction, and user interaction design. By optimizing for Oculus devices, the platform offers cultural explanations, dynamic interactions, and high-quality visual effects, providing a novel tourism experience.

This project focuses on creating a 3D virtual tour platform using 3D Gaussian Splatting (3DGS). The project highlights 3DGS's superior realism and speed, optimized rendering, and innovative UX approaches. Key features include AI-driven digital humans, customizable digital twin scenes, and first-person virtual tours. The system demonstrates robust performance and potential applications in enhancing digital interactions and immersive experiences.

In the emerging Metaverse era, this project revolutionizes exhibition booths with customizable virtual spaces. It features three main subsystems: the Metaverse Exhibition Scene for realistic user engagement, AI NPCs as interactive virtual assistants, and a Booth Customization System for designing personalized virtual stands. MetaBooth reduces costs, offers design flexibility, and reaches a global audience without physical constraints, setting a new standard for virtual exhibitions through advanced 3D modeling and real-time rendering.

This project developed with Unity and adapted for VR, offering an immersive historical experience. It highlights VR's advantages over traditional museums in terms of exhibition experience, audience engagement, and showcasing web3.0. The article covers design ideas, technical challenges, and solutions, and future improvements such as more interactivity and web3.0 integration, showcasing the museum as a model for digital museum development.

This project aims to enhance primary school national education using Augmented Reality (AR) technology. By employing Cognitivism, we make learning engaging and comprehensible, effectively forming concrete knowledge in students' minds. Following the Cone of Learning principles, it integrated an AI tutor for interactive reinforcement. Despite technical challenges, including AI language model API integration, this project successfully creates an immersive and interactive educational environment that significantly enhances national education learning.

This project integrates machine learning with industrial applications, focusing on text-to-image generation using techniques like DreamGaussian and Dreambooth. It targets industries, education, and entertainment, creating images from textual descriptions. The project includes system design, user experience optimization, and application in gaming, education, and virtual reality, demonstrating machine learning's practical potential across various sectors.

This project develops a Web 3.0 Unity3D game, in collaboration with YAKOO and ISHK, to educate young children about marine conservation in Hong Kong. The game features three levels focused on fish species, marine litter collection, and conservation questions. Successful completion rewards players with NFTs, enhancing learning and engagement.

This project develops an AR chemistry lab app for Hong Kong's Senior Secondary Chemistry syllabus. It offers experimental opportunities, realistic lab experiences, a safety quiz, periodic table, learning progress tracking, and personalized information. It uses UI design, object modeling, AR integration, and game process control for comprehensive chemistry learning.

This project aims to create an immersive VR fitness game, utilizing avatars for participatory exercise in a virtual world, combining fitness goals with fun. It focuses on building a realistic virtual environment using VR, incorporating MR to reduce Cybersickness, and integrating exercise into game tasks and challenges. Additionally, it offers correct fitness education through a practice mode that provides multi-perspective views of proper movements and explanations.

This project explores integrating 3D hair-changing systems, AI-driven recommendation systems, and augmented reality (AR) for innovative hairstyle customization. It examines AI's role in personalized hair advice and AR for hair overlay. Detailed system architecture, including 3D modeling, AI frameworks, and AR tools, is discussed. The report also covers hair adaptation with face reconstruction, AI-based recommendations, and real-time AR tracking. A user study evaluates the system's interaction and experience, aiming to create a powerful, user-friendly hairdressing solution using advanced technology.

This project explores the use of 3DGS technology to create realistic game scenes, enhancing player immersion at a lower resource cost. By developing a realistic-looking game focused on teen mental health, it aims to promote awareness and empathy, encouraging understanding and treatment of mental health issues..

In recent years, VR games have rapidly developed in the entertainment industry, offering immersive experiences highly sought after by players. However, VR games require higher resolution than standard games, making it challenging to improve rendering quality while maintaining fluency. Escape room games are also popular but face issues such as high costs, space limitations, and technical constraints. Converting escape room games to VR could solve these problems, but traditional VR rendering often results in blurry or cartoon-like graphics. The rise of 3D Gaussian Splatting technology changes this, enabling quick, realistic reconstruction of real objects and environments in VR, enhancing immersion and efficiency.

This project aims to revitalize Hung Hom by creating a sustainable community space atop the existing toll plaza. It addresses the area's environmental and urban challenges. Utilizing the Metaverse, the project promotes community engagement and awareness through innovative virtual experiences, allowing residents to explore the future Green Deck in a 3D environment. This approach demonstrates the benefits of green public spaces, encouraging local participation in urban renewal efforts and fostering a sense of involvement and ownership, ultimately creating a more sustainable and livable community in Hung Hom.

This project examines the use of BERT and RoBERTa_CN models for sentiment analysis of Chinese hotel reviews, integrating CNNs and RNNs to improve performance. Findings show limited efficacy in the Chinese context. The research aims to optimize hotel operations based on sentiment analysis, using machine learning and fuzzy string matching to classify reviews. This approach provides nuanced insights for targeted improvements, leveraging advanced NLP technologies to enhance service delivery and customer satisfaction in the hospitality industry.

This project aims to develop an intelligent guidance system integrated with traffic signals to improve safety for visually impaired individuals by detecting vehicle speed and proximity. Utilizing YOLOv5 for real-time obstacle detection, the system issues alarms when necessary and enters low-power mode otherwise. It enhances mobility and social inclusion with aviation-grade stability, weighing 247 grams, and a 15-day battery life. The paper evaluates smart guidance canes and robotic guide dogs, discussing their potential and limitations, and demonstrates the system's practical application in complex environments.

​This project aims to develop a novel 3D inpainting architecture using 3D Gaussian Splatting by distilling a 2D diffusion model. Previous methods, which involved independent 2D drawings from multiple views supervised together, often led to 3D inconsistencies in geometry and appearance. This project addresses these issues by leveraging a 2D pretrained diffusion model for color and depth inpainting and applying SDS loss in latent and image space.