Neural-Style-Transfer

neural style transfer by eager execution
paper: https://arxiv.org/pdf/1508.06576v2.pdf

weights used content_weight=(2)(1e3) style_weight=1e-2 style normalized by a coefficient of 0.75, instead of (height^2)x(width^2) considerable style transfered in just 200 iterations by mentioned hyperparameters

layers used: content_layers = ['block5_conv2']

style_layers = ['block1_conv1', 'block2_conv1', 'block3_conv1' 'block4_conv1' 'block5_conv1' ] reference : https://medium.com/tensorflow/neural-style-transfer-creating-art-with-deep-learning-using-tf-keras-and-eager-execution-7d541ac31398

Mixed-Neural-Style-Transfer

Used three loss functions two for style image and one for content image and tried to minimize them simultaneously