Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

competitions

portfolio

Stellar Classification with Spectral Data

I contributed a Stellar information dataset on Kaggle through data cleaning. I also performed logistic regression using Absolute Magnitude and B-V Color Index features to predict whether a star is a giant or a dwarf.

Spam Email Detection with Machine Learning

We conducted a performance analysis on Spam Email Detection with machine learning methods from research papers. We compare the methodology and results between XGBoost, LSTMs, and GRUs by reimplementing the research papers’ methods. Moreover, we suggested a novel approach to improve the existing methods by reducing training parameters and thus increasing the training efficiency.

Paper Warfare Game

We developed an innovative game which inspired by collective primary school memories of Hong Kong teentagers playing paper-folded aircraft back in their childhood. This game contains at least 2 hours of quality gameplay content including story mode and versus mode.

PolyU Faculty of Engineering Robotics Team

I joined PolyU Robotics Team in my first year and retired after the third year playing in Robocon. I participated in ABU Robocon 2019, ABU Robocon 2020, and ABU Robocon 2021. Each year we were given 500k HKD funding to build 2 robots to play in ABU Robocon HK (Biggest inter-university robotics event in Asia & HK).

End-to-End Network Intrusion Detection System

Traditonal methods has to rely on feature engineering. Our method detects Network Intrusion only with few bytes. This is done by turning binary information into image and feed into a vision model.

Deep Learning Image-to-Image Translation

My honours project extensively studied the use of GAN-based methods in Image-to-image translation. The first application could apply the style of Chinese paintings to a scenery image, the second application could paint the color on black and white sketches, and the third application could convert a cat into a tiger inside an image.

Deep Learning Image Cartoonization

In this work we proposed a Generative Adversarial Network that could transform real-world photos into high-quality cartoon style images. Aiming to create anime background images, we proposed (1) a new dataset that help increasing quality of output, and (2) an edge loss that creates a brushstroke feeling in the output image. This project is supported by Undergraduate Research and Innovation Scheme (URIS) in PolyU.

Intelligent Painter: Picture Composition With Resampling Diffusion Model

We present an intelligent painter that generate a person’s imaginary scene in one go, given explicit hints. We propose a resampling strategy for Denoising Diffusion Probabilistic Model (DDPM) to intelligently compose harmonized scenery images by injecting explicit landmark inputs at specific locations. Preprint available on Arxiv.

Anime Visual Novel Background Diffusion

A Stable Diffusion model is fune-tuned to produce visual novel backgrounds with just a few prompts. It is designed to allow creators to quickly and easily generate custom backgrounds for their visual novels. Model available on Hugging Face space.

Reimplement Shader from the video game TF2

A Rasterization+Raytracing renderer toy. Implemented the shader of the game ‘Team Fortress 2’ using only C++, no OpenGL library functions are called. Explored the shadering techniques used in video games like Half Lambert, Phong Shading, Fresnel Reflectance, XToon and more.

projects

publications

talks

teaching