For my AI Image Colorization Web Application, I designed and trained a conditional GAN using a ResNet34-UNet generator and PatchGAN discriminator. I improved output quality by replacing transposed convolutions with bilinear interpolation, eliminating checkerboard artifacts.
I implemented a two-stage training pipeline on COCO 2017 and built an async FastAPI service with tiled inference for high-resolution images. I also developed a Vue 3 SPA with an HTML5 Canvas live preview for real-time color correction.
For my Meter Reading Recognition System internship project, I built a YOLOv8-based OCR pipeline for utility meter reading, with a Flask REST API and a Tkinter desktop GUI. I’m currently pursuing a Master’s degree in Computer Science at Ivan Franko National University of Lviv.
