atris' log

About Me

Computer Science Graduate | Aspiring Tech Enthusiast

I'm Atris Yeung, a computer science graduate at the University of Nottingham in 2024 with a passion for both drawing and programming. I enjoy combining my creative and technical skills to create innovative solutions.

Linkedin : Atris Yeung

Github : @atris-Ctrl

My Projects

Check out some of the exciting projects I've been working on in the technology field. >_0

# Flower Segmentation by Deep Learning

Tool(s) Used: Matlab // Year 2024

Flower image segmentation presents a unique challenge due to the complexity of their shapes and colors. I explored two approaches for this task: transfer learning and constructing a deep learning model from scratch using a U-Net architecture. My goal was to accurately segment flowers from their backgrounds by training on the Oxford Flower dataset. I employed transfer learning with pre-trained models such as DeepLab v3+ and GoogLeNet. MATLAB made this process convenient, offering built-in models and an easy-to-use preview of the architecture. However, a significant challenge I faced was the training time, primarily due to the limitations of my laptop’s processing power. I was unable to train models with more than three layers efficiently. Here are some of my results.

Best performing images

Worst performing images

Architecture of U-net

# GeoFence Tracker

Tool(s) Used: Andriod Studio, Java // Year 2023

Github Repo Link

A simple application that allows users to track their walking path, record their steps, mark locations using Google map API. It also features a statistics section where users can view their walking trends, with tracking available even when the app is not open.

# Cooking Recipe Chatbot - Granny

Tool(s) Used: Python // Year 2023

Sometimes it is hard to decide what to eat so I created a recipe chatbot system called Granny has various feature with Identity Management for handling user names, Recipe Provision for suggesting detailed recipes, Question Answering using a comprehensive dataset and Small Talk for engaging in casual conversation. The system preprocesses user queries by tokenizing, standardizing, and stemming text to improve understanding. It uses TF-IDF and an MLP Classifier with NLTK python packages for accurate intent detection, asking for user clarification if confidence is low.

Best performing images

Worst performing images

# Dancing Game - Dance For Me

Tool(s) Used: Unity, C# ,Python // Year 2022

Github Repo Link

Dance for Me is a group project and an interactive AR game where players copy our character Dellboi, a 3D model's dance moves and get instant feedback using Mediapipe. It's designed to get desk-bound computer scientists up and moving in University of Nottingham. I worked as a front-end lead :>!

Starting Scene with different goofy dance to welcome people to play

Song Selection Scene with various choregraphies

About Me

My Projects

# Flower Segmentation by Deep Learning

# GeoFence Tracker

# Cooking Recipe Chatbot - Granny

# Dancing Game - Dance For Me

hello

Education

Others

# Crocheting ❤

# Drawing ❤