Learning Simple String Motifs for Sketching

I developed this project for IFT6141 Reconnaissance des formes course.


This paper describes the development of first intuitions for solving a problem simply defined as “taking an input and generating similar outputs”. Results can be applied to any variety of drawing or texturing applications. The problem is investigated by using input strings consisting of 0s and 1s representing simple line sketches drawn with a pattern of black and white.

Development has started with a brute force generation of a graph that is then traversed by selecting edges randomly. Vector quantization and clustering algorithms were studied in a second step, and applied to a graph generation process. Then, our brute force approach was combined with a hidden Markov model for both graph generation and traversing.

Results of each approach have led us to explore new solutions to the problem, and helped us to understand the problem from multiple standpoints. It has fostered our appreciation of different solutions and to evaluate their pros and cons.

You can reach the project report here.