Data Augmentation with Image Diffusion Morphing as an Educational Tool
Edward Motoaki Wenngren, Tad Gonsalves · 2024
This paper goes into the viability of image diffusion morphing as a method of data augmentation. With the recent surge of AI and machine learning models, getting more data to achieve better results from models is imperative and there can never be too many methods for data augmentation. There is pre-existing research which delves into the viability for data augmentation using pure synthetic images, however, there is no research into the viability of using image diffusion morphing, a technique used to smoothly interpolate two images together using image diffusion as of yet. This paper hopes to introduce and investigate the viability of image diffusion morphing as a data augmentation method for future generations.