Projects - Yuk
Projects - Yuk
Program: TrafficLab3D
TrafficLab puts accessibility at the forefront, with just access to mp4 CCTV footage and knowing where that location is on Google Maps, anyone can create a fancy digital twin demo that demonstrates advanced computer vision, especially for students, individual investigators, and enthusiasts who might not have access to camera calibration and synchronized high quality satellite imagery.
Program: Text-to-SQL SLM
Trying out finetuning on language model with one of the easiest example and deployment on Hugging Face Spaces. Performed QLoRA on a Gemma 3 1B model using a text-to-SQL dataset. The finetuning and deployment itself is quite simple, so I focused more on the math and some of the best practices in this currently very shaky field.
Program: Computer Vision MAT3562E
My work for the MAT3562E - Computer Vision course at HUS - VNU. A Traffic Collision Anticipation Model with hybrid feature extraction using EfficientNetB0 and Farneback Optical Flow, feeding into a Temporal Convolutional Network (TCN) to achieve a continuous sliding window risk score prediction. Trained and inference on only dashcam footage, no additional sensory input.
Paper: ResearchM. MAT1204
My work for the MAT1204 - Research Methodology course at HUS - VNU. Throughout this project, I was able to pretty much grasp the entire LLM + integration/application field of research, from how expansive the landscape of LLMs actually is-to working with them and various integrations. I also had the chance to gauge my own capability considering I worked on this end-to-end all on my own. Nevertheless, this project also involved a lot of disillusionment and negativity.
Website: Yuk's Personal Blog
My little space to share my stuff and work publicly, I hope it can help and even inspire those who come across it. The website itself is built with pure HTML, CSS and JavaScript, then uses Jekyll to generate the dynamic content and posts. Hosting is currently done using GitHub Pages.
Báo cáo: Học máy MAT3533
Báo cáo hết môn Học máy MAT3533 tại HUS - VNU của nhóm mình: Ứng dụng các phương pháp học máy trong chủ đề đánh giá sức khỏe tâm lý. Tuy sản phẩm cuối cùng có hơi khác với mong đợi đầu của mình, đây vẫn là báo cáo học thuật chi tiết đầu tiên của mình.
Program: MNIST Workshop
This Java project implements a Feedforward Neural Network (FNN) for the MNIST dataset, providing a foundational exploration of machine learning. It allows users to create, train, test, and save FNN models, emphasizing core concepts like forward/backpropagation and gradient descent without external dependencies.
Program: Gomoku Bot
This Java project implements a classic Gomoku (5-in-a-row) game with AI opponents. Players take turns placing stones on a 19x19 board, aiming to be the first to form a line of five. The key feature is a custom-built heuristic minimax framework for the bots, offering valuable hands-on experience with GOFAI (Good Old-Fashioned AI).
Program: Chess Game
This Java project is a remnant of an ambitious attempt at creating a fully-fledged Chess Game app. Though I stopped working on it back then, Its still one of my most extensive work yet and marks a pivotal point in my coding journey. The app features many functionality with a hybrid between GUI and CLI and stops at an experimental chess bot that just plays random moves.