Leon Liu's Projects
Demo of 2D platformer controls and collectibles
This repo contains the code examples for the winter 2024 ACM Cloud AWS workshop series.
Uses NLP & LSTM to detect insincere Quora Questions
An AI Watch Party app that creates a highlight reel based on peak heart rate!
An implementation of Conway's Game of Life
Finduo is a gaming social media app designed to help gamers find their ideal duo
Use Transformers to predict Time-Series smartphone positioning from GNSS data.
Annotated implementation of Andrej Karpathy's GPT from scratch
Math Mage is a game developed from scratch without a game engine using Java's JSwing Application library. It is a wave-based top-down arena shooter where a math mage fights off hordes of monsters, undergoing trigonometry trials to gain boons in the form of exciting new weapons and power-ups. It is a fun way to learn your Trig Table for sin, cos, tan, csc, sec, and cot in degrees and radians! Hone your math skills and beat it as fast as you can in the hardest difficulty for a high score on the leaderboard. I created the UI and gameplay programming as well as art assets. The java project is hosted on Repl.it but can be downloaded and compiled to play on your computer.
Reverse Proxy built with Deno, UCLA ACM Cyber
Silly PVP RTS Tower Defense Game!
This is a web application that allows a user to sketch a doodle to search for a image. An AI model will recognize the drawing and retrieve images that match the query.
Implementation of Transformer from Scratch
Contains a module for functions that build and evaluate a U-Net Model which is a modified auto-encoder image-to-image architecture that includes skip connections from encoder to decoder layers with matching input dimensions to retain features from the encoded image. The model is designed for image segmentation and produces a mask based off x-ray data. The model is used with lung x-rays to segment out lungs, brain scans to segment out brain tumors, and pictures of a room to segment out a human figure. With the lungs dataset, it is able to manage 98% accuracy with a wide range of 3,4,5 encoding block of Conv2D and MaxPooling and BatchNormalization + 1 latent layer without pooling and the mirrored number of decoding layers that do Conv2DTranspose for deconvolution and UpScaling to reverse MaxPooling making 7,9,11 layers with doubling filters/kernels in each encoding block from 16,32,64,80 that double with each block like in a traditional CNN and then half with each decoding block.