Released in 2016, this net automatically colorizes a grayscale image, exploiting a combination of local and global image features. Local features are extracted in a fully convolutional fashion, while the extraction of global features was developed leveraging the labels of the ImageNet Competition dataset during training.
Number of layers: 62 |
Parameter count: 44,457,314 |
Trained size: 178 MB
Examples
Resource retrieval
Retrieve the resource object:
Out[1]= |  |
Get the pre-trained net:
Out[2]= |  |
Basic usage
This net takes a grayscale image as input and outputs the A and B channels in the LAB color space. Currently it needs an evaluation function to merge its output with the luminance of the input:
Colorize a grayscale image:
Out[4]= |  |
Performance evaluation
Get a color image:
Compare the colorization performed by the net with the ground truth:
Out[6]= |  |
Export to MXNet
Export the net into a format that can be opened in MXNet:
Out[7]= |  |
Export also creates a net.params file containing parameters:
Out[8]= |  |
Get the size of the parameter file:
Out[9]= |  |
The size is similar to the byte count of the resource object:
Out[10]= |  |
Represent the MXNet net as a graph:
Out[11]= |  |
Requirements
Wolfram Language 11.2
(September 2017) or above
Reference