ivadomed/model-canal-seg

Test the model's latest version on sct-test-large

Closed this issue · 1 comments

It would be good to evaluate the model's performance on sct-test-large dataset to see how it generalizes on various types of data.

I could look one by one at the image, generating a qc report after the inference, but I was wondering if there was some metrics to measure the accuracy of the model on all of this data that don't have groundtruth...

If you don't have a ground truth, more complicated to add metrics, visual assessment is all you could do I think