Generative Adversarial Network - Fake Houses
Through implementation of a Generative Adversarial Network (GAN) model to create synthetic thumbnail images of houses. This was a real-world problem completed as a capstone project for UBC Master of Data Science program 2 years ago. My target was to recreate the results and simulate the problem by creating a house images dataset through web scraping and to complete the project in two weeks rather than the original capstone timeline of two months.
Tools used: Pytorch, web scraping, Google Cloud Platform, Selenium
Link to the project
Semantic Image Segmentation
In a second phase of my Neural Style Transfer (NST) project, I have applied semantic image segmentation to exclude person figures from the style transfer. This project deploys a U-Net deep learning architecture capable of separating person figures from their background. The image segmentation pipeline is then stacked on the NST pipeline to augment the output image.
Tools used: Tensorflow, Google Cloud Platform, data augmentation, data loading and image preprocessing
Link to the project post
Link to the project
Neural Style Transfer
This is a personal project on implementation of Neural Style Transfer, an unsupervised machine learning model. The goal of this project was to read and understand a research paper (Gatys et al. (2015)) and implement a Tensorflow code that would run on a GPU-accelerated machine (an NVIDIA product). Skills learned also include image processing and transformation, Transfer Learning and building an NVIDIA docker image using Podman.
Tools used: Tensorflow, Keras, PIL, NVIDIA Jetson Nano, Podman, QEMU
Link to the project post
Link to the project
Canadian Heritage Funding
This is a group project on a classification problem demonstrating my teamwork performance and knowledge of Data Science workflows, Github Actions, creating Docker images and formulating a reproducible analysis, model, and report.
Link to the project
COVID-19 Testing – Online Booking Website
This is a proof-of-concept website designed at the beginning of the pandemic to reduce the pressure on Public Health call centers in Calgary, Alberta caused by higher than normal volumes related to COVID-19. The website serves three main purposes: Screening of the patients, booking a time for walk-in or drive thru testing appointments and storing the data in a SQL database, and modifying booked appointments.
Languages used: Python, Flask, SQL, HTML, JavaScript, CSS
Link to the project
Demo