Color palette generation with Conditional Generative Adversarial Networks

BURYKIN, SERGEI (author), Sadikov, Aleksander (mentor)

Creation of unique color palettes is a challenging task for designers all around the world. Every year it becomes increasingly difficult to create new color palettes. Color theory describes different algorithms for creation of color palettes, but these algorithms limit the possible color combinations. Designers all around the world are trying to find new ways to complete this task. We are trying to apply deep learning algorithms in order to create the color palettes which were never seen before. Using Generative Adversarial Networks (GAN) we can expand the amount of unique color palettes, since GANs are not limited by conventional algorithms.

• Language: English
• Keywords: Artificial Intelligence, Deep Learning, Generative Adversarial Networks
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2021
• PID: 20.500.12556/RUL-128188
• COBISS.SI-ID: 69404675
• Publication date in RUL: 05.07.2021

Secondary language

• Language: Slovenian
• Title: Ustvarjanje barvne palete s pogojnimi generativnimi nasprotniškimi mrežami

Abstract

Ustvarjanje edinstvenih barvnih palet je zahtevna naloga za oblikovalce po vsem svetu. Vsako leto je vse težje ustvarjati nove barvne palete. Teorija barv opisuje različne algoritme za ustvarjanje barvnih palet, vendar ti algoritmi omejujejo možne kombinacije barv. Oblikovalci po vsem svetu poskušajo najti nove načine za dokončanje te naloge. Poskušamo uporabiti algoritme globokega učenja, da bi ustvarili barvne palete, ki jih še nikoli nismo videli. Z uporabo Generativnih Nasprotniških Mrež (GAN) lahko razširimo količino unikatnih barvnih palet, saj GAN-i niso omejeni z običajnimi algoritmi.

• Keywords: Umetna Inteligenca, Poglobljeno Učenje, Generativne Nasprotniške Mreže