Skip to content

Using DCGAN to generate images of nature (mountains, flowers, etc.)

License

Notifications You must be signed in to change notification settings

Skuldur/DCGAN-Nature

Repository files navigation

DCGAN-Nature

A DCGAN written in Keras based on https://arxiv.org/pdf/1511.06434.pdf and trained using a custom mountain dataset (not included) and the Oxford Flower Dataset http://www.robots.ox.ac.uk/~vgg/data/flowers/

Examples

Examples of images generated using the mountains_generator and flowers_generator weighs

alt text alt text

How to use

The command line interface for the DCGAN is as follows:

--mode             choose between the two modes: (train, generate)
--batch_size       The size of each batch (default: 128)
--learning_rate    The learning rate for the Adam optimizers (default: 0.0002)
--beta_1           The beta 1 value for the Adam optimizers (default: 0.5)
--epochs           The amount of epochs the network should train (default: 100)
--data_path        The path to the images that should be used for training
--model            Choose which weights to use for the generation (default: mountains)

Example

python dcgan.py --mode train --data_path images/small_flowers

About

Using DCGAN to generate images of nature (mountains, flowers, etc.)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages