Bring your Art to Life with Pix2Pix

PIX2PIX — Machine Learning Tool for Artists

Pix2pix can also be used to:

Encoder -> Decoder

The Generator’s Architecture

The Generator network is made up of these two networks:

The Generator’s Encoder Architecture

The Generator’s Decoder Architecture

There are six skip-connections in a Generator network. The concatenation happens along the channel axis:

Discriminator’s Architecture

Figure 1 PatchGAN Architecture with 5 convolutional Network

GAN’s Architecture

Figure 2 GAN’s Architecture

Pix2Pix Network’s Training

Following are the steps that involve training the model for the Pix2Pix algorithm:

1. Import TensorFlow and required Libraries

2. Load the Dataset

3. Input Pipeline

4. Build the Generator

5. Generator loss

6. Build the Discriminator

7. Discriminator loss

8. Define the Optimizers and Checkpoint-saver

9. Generate Images

10. Training

The Training Loop:

A sketch of a building
Real Image of a building
A sketch of a shoe -> An image of a shoe
A sketch of a purse An image of a purse
A diagram of Venice
Aerial view of Venice

Conclusion

Reference

Affine is a provider of analytics solutions, working with global organizations solving their strategic and day to day business problems www.affineanalytics.com