CODEWITHSUNDEEP
classificationresnet
em_
em_

Reputation: 1

Resnet18 model returning 0% accuracy on test set

I have trained a Resnet18 PyTorch model on a set of images for binary classification. I pre-loaded the Resnet18 architecture and trained the model for 10 epochs and a batch size of 4. My training accuracy by the end of training is pretty high (likely overfitting as I only have 400 images in my training set and no validation set). However, when I come to test on my test set (of 100 images), I get a 0% accuracy. I believe this is wrong as with a binary classifier, the lowest/worst performance should technically be 50%. Can anyone explain what may cause a 0% accuracy?

Thank you in advance.

Upvotes: 0

Views: 54

Answers (1)

ludakris
ludakris

Reputation: 1

You probably want to double check that you are computing/printing the test accuracy correctly, as you mentioned even random guessing should get 50% (assuming labels are evenly distributed).

Upvotes: 0

Related Questions