Elderly dementia treatment using ai-generated image content

Abstract

In this project, we propose two approaches and talk about our ideas for each achievement, as well as the advantages and disadvantages of them. For the first approach, we want to modify the architecture of an existing text-to-image model and to make our own (text and image)-to-image version and then train it from scratch. Although the results of this approach are not yet satisfied, there is more potential and future work to be addressed. Meanwhile, the second approach is about combining two existing models (text-to-image and image-to-image) to create an early version of a working application to serve our purpose. This proves to be a more feasible way and has already produced acceptable results.