Розробка застосунку для часткової аугментації зображень на основі генеративних моделей

Abstract

Data augmentation is a crucial step in creating machine learning and deep learning models. Existing tools and frameworks lack the functionality to augment the target class and specific image regions effectively. In this work, we present a web application aimed at facilitating and semi-automating the image augmentation process, focusing on transforming the target class for segmentation tasks. The application demonstrates its effectiveness when tested on a specific dataset and offers flexibility for integrating new augmentation models for images from different domains, using app modular architecture.