De-sketching

Peer reviewed: 
Yes, item is peer reviewed.
Scholarly level: 
Undergraduate student
Final version published as: 

L. Bragilevsky and I. V. Bajic, "De-sketching," IEEE Multimedia Signal Processing Workshop (MMSP), Vancouver, BC, Aug. 2018.

Date created: 
2018-06-22
Keywords: 
de-sketching
deep learning
Abstract: 

Many software applications exist for plotting graphs of mathematical functions, yet there are none (to our knowledge) that perform the inverse operation - estimating mathematical expressions from graphs. Since plotting graphs (especially by hand) is often referred to as "sketching," we refer to the inverse operation as "de-sketching." As the number of mathematical expressions that approximate a given curve can be quite large, in this demo we restrict our attention to polynomials, and present a deep model that performs de-sketching by finding the best second-degree polynomial to fit the curve in the input image. Currently, our trained model is able to provide reasonably accurate estimates of polynomial coefficients for both synthetically-generated and hand-drawn curves.

Language: 
English
Document type: 
Conference presentation
File(s): 
Statistics: