What is the standard way of converting grayscale image to RGB to make it work with RGB supported CNN models?

89 views Asked by At

As I found, most CNN backbones (resnet, convnext, inception, efficientnet etc.) work in RGB. I have a dataset with greyscale images. My question is: What is the standard way to convert greyscale to RGB inorder to make it work with RGB backed CNN models available?

What I am currently doing is: adapter = tf.keras.layers.Conv2D(3 ,1) that essentially takes 1 channel as input and outputs 3 channels with a kernel size of 1.

An alternative way can be to simply copy the BL value (say 140) to RGB (140,140,140). Are there better approaches that I already mentioned?

0

There are 0 answers