Login
       
  • Machine Learning for Creative Transfer

Dare, Eleanor, 2019, Conference or Workshop, Machine Learning for Creative Transfer at Hong Kong Poly University Artificial Intelligence Symposium, Hong Kong, 26-Feb-2019.

Abstract or Description:

The first algorithm for neural style transfer was proposed by Gatys et al (2015), since then, Style Transfer for images using Computer Vision has become increasingly efficient. Convolutional neural networks (CNN) are the main technique which make style transfer possible. Via Deep learning CNN networks deploy a process of repeated optimisation until one image adopts the style of another while retaining its own original content. The potential for style transfer in Quality Control is worth exploring, in particular, the development of visual, haptic and olfactory style transfer, to measure quality within a multi-modal, multi-dimensional problem space. Most quality control for garments is currently undertaken by human workers, with significant potential for errors to go unchecked. Such work is also tedious and fatiguing, making quality control an obvious, if challenging area for development, benefiting both industry and individual workers. Style transfer is a relatively new technique, one which has not yet been deployed within the garment industry, or within a multi-modal form.

Subjects: Other > Mathematical and Computer Sciences > G700 Artificial Intelligence > G740 Computer Vision
School or Centre: School of Communication
Funders: Hong Kong Poly University
Date Deposited: 26 Mar 2019 13:33
Last Modified: 26 Mar 2019 13:33
URI: http://researchonline.rca.ac.uk/id/eprint/3843

Actions (login required)

View Item View Item