Information-theoretic stochastic contrastive conditional GAN: InfoSCC-GAN. Paper accepted to NIPS 2021 BDL Workshop
Generation of data on discontinuous manifolds via continuous stochastic non-invertible networks. Paper accepted to NIPS 2021 BDL Workshop
Turbo-Sim: a generalised generative model with a physical latent space. Paper accepted to NIPS 2021 Machine Learning and the Physical Sciences Workshop
Self-supervised contrastive learning method with pretext task regularization for small-scale datasets. 🏆 SOTA on CIFAR20 Unsupervised Image Classification. Paper accepted to ICCV 2021 2nd Visual Inductive Priors for Data-Efficient Deep Learning Workshop
MITACS Globalink Internship. I was working on project Spatial Clustering and Possibility Areas for Maritime Search and Rescue Operations Optimization
I was working on projects related to analysis of spatial greenhouse gas emissions and nightlight light data in particular on mitigating geolocation errors in nighttime light satellite data and global CO2 emission gridded data.
I was using Computer Vision, Machine Learning and Deep Learning to solve real-worl problems in various industries:
Computer Vision Internship. During the internship, I have created two projects:
Scikit-Learn; Keras; Pytorch; OpenCV; DLib; Tesseract
QGIS; Google Earth Engine
Ukrainian (Native); English (Advanced); Russian (Advanced); French (Beginner); Polish (Beginner); German (Beginner)
Web libraries: Flask; Operational systems: Linux, Windows; Git, SVN, Docker
Application for predicting air quality index (AQI) based on satellite data and on-ground observations using machine learning. This project was local round winner in Lviv, Ukraine.
Application for analyzing raster images from satellite, detecting trees and green zones in cities, calculating the percentage of green zones in cities using computer vision techniques. This project was local round winner in Lviv, Ukraine.