Pre-trained Object Detection API model for greyscale images (1 channel)

Question

I want to detect features, and possibly segmented them, in CT scan images (greyscale images) by means of the Object Detection API from Tensorflow. So I have two questions:

-Is there any pre-trained model within the Object Detection API that takes greyscale images as inputs for fine tuning training?

-Or, How can I train exisiting models, that are for RGB images, so that I can train them with greyscale images? Which modifications should be made in the code?

Answer 1

Just in case someone would be interested in how I tackled this problem:

For segmentation of grayscale images, I found out this architecture: U-Net The tensorflow adaptation by Akeret et al. can be found in Github - UNet

I did not use API Object detection in the end since I was not sure how well the pre-trained models would perform in 1-Channel Images since they have been created for RGB-images.