Resnet50: ValueError: Error when checking input: expected input_2 to have shape (256, 256, 3) but got array with shape (256, 256, 1)

I am trying to train Resnet50 with grayscale images e.g Channel = 1. while using fit_generator in keras I am facing a strange error of channel cross checking. I think the Resnet50 model is trying to use channel of 3 instead of 1. I asked in keras github issues for the same. Any suggestions is highly welcome.

Link for github issue:-

This forum is for the current pytorch version of the course. For help with keras, you should use the old #part1 category.

Oops. Sorry!!