Tensorflow implementation of the paper: DualGAN: Unsupervised Dual Learning for Image-to-Image Translation
Link to Complete Dataset
- datasets : Contains 2 datasets namely "facades.zip" and "day-night.zip" which were used to train the model.
- results : Contains the results obtained from our implementation for each of the above dataset.
2.1 Receptive fields used for the day-night dataset: 70x70, 16x16, 1x1
2.2 Receptive fields used for the facades dataset: 70x70
- training_parameters.json : Contains the parameters for training the model such as no. of epochs, dataset_name, batch size, directory paths.
- tf1_dualgan.py : File containing the core dual GAN architecture and the whole logic for generators, discriminators, training, testing.
- helper.py : Contains the Dataset class with helper functions such as loading and fetching images.
- calc_receptive_field.py : Contains logic for calculating patch sizes for different receptive field sizes.