Artificial intelligence meets art

• Main Author: Yeong, Linus Ka Shing
• Other Authors: Zheng Jianmin
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Artificial intelligence; Art; Doodle; Sketch
• Online Access: https://hdl.handle.net/10356/175207
• Institution: Nanyang Technological University

With the introduction of Artificial Intelligence (AI) in the art domain, there have been huge advancements ranging from image editing to the creation of digital art. One of the most common first interactions with art for young children is drawing, and those with the passion for drawing will gradually learn to create more detailed sketches and develop their art style. This project develops an educational art tool called Doodle-to-Sketch, which is made with the intention to allow for young kids to acclimatise with sketching by transforming their doodles into a sketch. This tool is also made to be cost effective, making use of open-source technology, pre-trained models as well as custom models to create an inexpensive solution which can be used for educational purposes. The main process involves training a doodle classifier and using it to classify the doodle. The predicted label is then incorporated into a prompt and used as an input to a pre-trained AI image generator to output realistic images which aims to be as close to the doodle as possible. Finally, image processing is applied on the generated images to obtain sketches for the young children to refer to. This research is part of the project “AI in Art Studio Creating image-based artworks using machine learning and science data” under NISTH co:Labs Grant.