This is an unofficial PyTorch implementation of the paper Image-to-Image Translation with Conditional Adversarial Nets.

If you find this code useful, please star the repository.

• Clone this repository

git clone "https://github.com/FarnoushRJ/MLProject_Pix2Pix.git"

• Install the requirements

  Other Requirements

  • Pillow 7.0.0
  • numpy 1.18.4
  • matplotlib 3.2.1
  • barbar 0.2.1
  • torch 1.5.0
  • torchvision 0.6.0

• Facades and Maps datasets can be downloaded from this link.

Data Directory Structure

|__ DATASET_ROOT
    |__ train       
    |__ test
    |__ val     

cd train/
python train.py --args

The models is trained for 200 epochs on both Facades and Maps datasets.

Facades Training Loss	Maps(AtoB) Training Loss	Maps(BtoA) Training Loss

Input, Fake Target, Real Target

Input, Fake Target, Real Target (AtoB)

Input, Fake Target, Real Target (BtoA)

• Models
  • Modified Model for deblurring, denoising and Inpainting

• pytorch-CycleGAN-and-pix2pix
• pytorch-pix2pix