Image-colorization-GAN

Description

This Project is implementation of image colorization from black and white images using GAN.

Libraries

• TensorFlow
• sklearn
• PIL
• matplotlib
• numpy

Dataset

The dataset used to train this model is : https://www.kaggle.com/datasets/jessicali9530/stl10/data

Custom dataset

To use custom dataset ensure you create a folder data which has all the coloured images.

Training

Config

batch_size = 4
img_size = 120
dataset_split = 2500
master_dir = 'data/'

Epochs: 200

Total training time: 6867.23 seconds

Using model

To use the model you can directly use the model from the repo generator.keras which outputs colourized image when black and white image is provided.

Example:

generator = tf.keras.models.load_model('generator.keras')
# Generate output for the test_x data
y = generator( test_x[0:10] ).numpy()