Skip to content

This project is developed to demonstrate DCGAN network architecture which generates abstract art images.

License

Notifications You must be signed in to change notification settings

niraj2477/GAN-Abstract-Art

Repository files navigation

ABSTRACT ART GENERATOR


This project is developed to demonstrate **DCGAN** network architecture which generates abstract art images.

Table of Contents

Introduction

Thsi project has implemented DCGAN network architecture which contains Generator and Discriminator model to generate sample of size (512 _ 512 _ 3) based on given training data.


Architecture

It is composed of Generator and Discriminator model, To understand further..

The Generator model implements transpose convoluational layer to perform convoluation as well as upsample the output of convoluation using strides of (2,2) whicle keeping the padding same and applying leakyRelu activation to each genrated output. Graphical representation is as follows: Generator


The Discriminator model implements convoluation layer to perform convoluation and apply maxpooling of (2,2) to downsample images and afterwards uses fullyconnected layer to output prediction by applying sigmoid activation. Graphical representation is as follows:

discriminator


Training & Evaluation

Following parameter were used while training the network

 Image height : 512

 Image width : 512

 Image Channel : 3

 batch size : 16

 latent dim : 200

 number of epochs : 200

network seems to generate close to real images out of provided latent dimension which can be seen during the training.

some of the sample that were generated by network during testing,

further model can be trained for more epoches to get better result from the network.

References

dataset link

https://www.kaggle.com/datasets/greg115/abstract-art

other useful links

https://github.com/bnsreenu/python_for_microscopists/blob/master/248_keras_implementation_of_GAN/248-cifar_GAN.py

https://github.com/nikhilroxtomar/DCGAN-on-Anime-Faces/blob/master/gan.py

About

This project is developed to demonstrate DCGAN network architecture which generates abstract art images.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages