validation or test accuracy

[DANNALI35]

Anyone knows about the validation or test accuracy of this model?

I implement the paper using keras and use 10% images of each category as validation data, it turns out the model performs outstanding on training data while the accuracy on validation data fluctuates wildly, from 10%~ 99%, quite strange.

The paper only mentions on accuracy on all images, this can be similar to training accuracy, but the high training accuracy can be a result of overfit.

[prodanovann]

Hi, a bit late answer, but for whoever might read this: If you take 10% of the images you have the same patients in train and test and the performance estimation is not valid.