Deep Learning in Mobile Sketch-Based Modelling

Patrick Werner, Wolfgang Heiden, Ernst Kruijff, André Hinkenjann
International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), page 1-5 - oct 2021

Abstract

In this paper, performance of different deep learning methods is evaluated to improve sketch based modelling in the popular MagicToon mobile application, through their usage in the classification, contour detection, segmentation and automatic rigging stages. Results show that classification of browser-drawn sketches can be generalized to real life drawings. In comparison to previously proposed methods, detection performance for real drawings was improved using contour detection. Based on the results future improvements of sketch-based modelling using deep learning can be expected.

Images and movies

 

BibTex references

@InProceedings{WHKH21,
  author       = {Werner, Patrick and Heiden, Wolfgang and Kruijff, Ernst and Hinkenjann, Andr{\'e}},
  title        = {Deep Learning in Mobile Sketch-Based Modelling},
  booktitle    = {International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME)},
  pages        = {1-5},
  month        = {oct},
  year         = {2021},
  publisher    = {IEEE},
}

Other publications in the database

» Wolfgang Heiden
» Ernst Kruijff
» André Hinkenjann