Project - Value of colour in deep convolutional neural networks

Funding resources

Evropská unie - European Cooperation in Science and Technology (COST)

- whole funder (not assigned - 2024-09-29)

On the project

The current approach to the design and implementation of computer vision systems based on deep convolutional networks leads to a loss of information included in colour. Our goal is to minimize the loss, and consequently, to improve performance of the systems. The improvement will allow application of colour computer vision in tasks, where its application has not been possible so far.

Keywords
Digital images, electromagnetic radiation, spectral band, color digital images

Mark

CA19145

Default language

English

People responsible

Škrabánek Pavel, doc. Ing., Ph.D. - principal person responsible

Units

Institute of Automation and Computer Science
- (2021-07-01 - not assigned)

Responsibility: Škrabánek Pavel, doc. Ing., Ph.D.

VUT

Faculties

University Institutes

Parts

Value of colour in deep convolutional neural networks