HennaGAN: Henna Art Design Generation using Deep Convolutional Generative Adversarial Network (DCGAN)
Sayeda Samia Nasrin, Risul Islam Rasel
Abstract
The quest of imparting intelligence into human creations have led to the inception and progress of Artificial Intelligence (AI). One application of AI is the generation of things like text or images, using a class of neural networks known as Generative Adversarial Networks (GANs). Nowadays, AI can generate different sorts of artwork and arts using GANs. However, many traditional art design patterns that are tied historically to a particular segment of the demographics and geographic are left unexplored so far. Henna artwork is one of them. In this regard, this paper aims at the generation of Henna design patterns, which is a widely popular artwork involving of complex creative designs applied mostly on the hands as a type of temporary tattoo in the Indian sub-continent and parts of Asia. The HennaGan introduced in this paper shows that Deep Convolutional Neural Network (DCGANs) can be used to generate henna design images with variations effectively. It also creates the base for research into creative Henna art generation using DCGANs and provides insights into how the network parameters can be tuned to obtain a good result.