Wu S.-JLin P.-SHuang P.-CSHOU-DE LIN2021-09-022021-09-022020https://www.scopus.com/inward/record.uri?eid=2-s2.0-85104626309&doi=10.1145%2f3447490.3447493&partnerID=40&md5=db83173c2b78ea72e98c680217088122https://scholars.lib.ntu.edu.tw/handle/123456789/581447Sketching is a type of artistic skills that most people can acquire during their childhood. In this paper, we present a web-based GUI environment which interactively teaches users how to sketch. Based on neural-network variational autoencoder (VAE), the environment uses generated strokes to guide users, however, the drawing quality cannot naively meet standards required for tutoring purposes. We conducted qualitative analysis, investigated the influence of latent variable on the drawing quality, and choose transformer as decoder. The proposed technique enhances the interactive drawing quality, and has the potential to build a sketch-tutoring software for young children. ? 2020 ACM.Artificial intelligence; Interactive; Neural network; Sketch[SDGs]SDG3[SDGs]SDG4Learning systems; Auto encoders; Interactive sketch; Latent variable; Qualitative analysis; Web based; Young children; Artificial intelligenceconference paper10.1145/3447490.34474932-s2.0-85104626309