Pix2pix for Aerial-Map dataset

Implementation of pix2pix methods to pass Aerial images to maps, the algorithms that we implemented are:

• Unet [supervised]
• GAN [unsupervised]
• L-GAN [supervised]
• CycleGAN [unsupervised]
• L-CycleGAN [supervised]

Setup

To install all required dependencies run:

pip install -r requirements.txt

Running

To start training and get output with:

python train.py -m=<model> -e=<epochs> -d=<dataset>

Possible values for parameter model are: unet,gan,lgan,cyclegan,and lcyclegan.

Possible values for parameter d are: a2m and inria

epochs is the number of epoches that you want to run

All the output images are saved in the /output directory

You can use tensorboard in the /logs directory to visualize the loss