Extending DenseNet121 and MedNist Classifier Tutorial for 3-Channel Images

[DocandBean]

Hi MONAI Community,

In trying to extend the code from the MedNist Classifier tutorial to use 3-channel jpgs, I receive an error message regarding the input size:

RuntimeError: Expected 3D (unbatched) or 4D (batched) input to conv2d, but got input of size: [300, 1, 3, 256, 256]

The change I expected to use is the 'in-channels' from 1 in the tutorial to 3 as follows:

model = DenseNet121(spatial_dims=2,in_channels=3,out_channels=num_class).to(device)

The input size in the MedNist Classifier tutorial is [300, 1, 64, 64] and thus changing the in_channels alone is insufficient for 3-channel jpgs. Is there an argument to LoadImage to flatten? Or another approach?

Any advice would be much appreciated.

[DocandBean]

I missed the AddChannel() in the transforms. Problem solved. 🙏

Thanks all!