Segmentation when Image Channels Are More Than 3

I have Recently Started Working on a Dataset For Image Segmentation.But The Problem that I am encountering is that That isnt of having just 3 Channels,I have been given 7 images Each Corresponding to A different Channel for one Segmentation Mask and I am confused how to get them Ready For The Model.
Could @jeremy you Please point me in The right direction

1 Like

I assume you have 7 channel input image.

You can call Resnet inside Unet learner.

Could you please try with changing the ResNet Input channel to 7 instead of 3 and try loading it in learner.
It might work for you, please keep me posted.

Last week I changed the stride for restnet to 1 instead of default 2 for small object on image never tried with channels though.

Does the image has 7 channels or for each image you have 7 different images corresponding to a different segmentation mask for that image?

Here you can find a starter the code to:

  • load multi channel image
  • adapt resent first conv-layer to accept more than 3 channels