Colourizing black and white images, using pix2pix gan architecture.
Used flicker8k dataset.
- 6k training set
- 2k testing set Transformed the images in the dataset to L*ab colour format and used L (lightness) channel as input and ab (colour) channel as output.
Trained the model for 100 epochs.
1st row is the input (black and white image), 2nd row is the predicted image and 3rd row is the original image.