Content Aware Image Recolorization

A deep neural network which recolors an image according to a given target colour that is useful to express images in various colour concepts. The network is capable of performing a content-aware recolorization based on the target palette. The main features are:

• A convolutional encoder that extracts features from the input image
• A convolutional decoder that generates a colourized output image
• A perceptual loss function that measures the similarity between the output and the ground truth images
• A dataset loader that can handle different image formats and sizes

Results

a) The source image

b) The Target Palette

c) The recolorized image as per the target palette

Usage

Run DL2022/run.py as

python run.py -i 'path_to_input_image' -p 'hexcodes_of_color_palette' -m 'path_to_saved_model' -o 'path_of_output'

Example

python run.py -i 'Samples/3/3.png' -p '#507b71' '#6caebc' '#6ead9c' '#afd9c3' '#b8dfdc' '#ecebd7' -m 'saved_model/model.pth' -o 'output.jpg'

Refrence

PaletteNet: Image Recolorization with Given Color Palette