CODEWITHSUNDEEP
pytorchconv-neural-network
Sohail
Sohail

Reputation: 43

How to add a channel layer for a 3D image for 3D CNN

I am working on a medical dataset of 3d images the shapeimage.shape gives me (512,512,241) I assume it is height , width, depth but when I run 3D cnn i get this error Given groups=1, weight of size [32, 3, 3, 3, 3], expected input[1, 1, 241, 512, 512] to have 3 channels, but got 1 channels instead. what should be done? I am using pytorch.

Thank you in advance.

Upvotes: 1

Views: 401

Answers (1)

Shai
Shai

Reputation: 114786

Your network expects each slice of the 3d volume to have three channels (RGB). You can simply convert grayscale (single channel) data to "fake" RGB by duplicating the single channel you have:

x_gray = ...  # your tensor of shape batch-1-241-512-512
x_fake_rgb = x_gray.expand(-1, 3, -1, -1, -1)

See expand for more details.

Upvotes: 1

Related Questions