Does anyone know why does the accuracy for the validation set is starting from a high point instead of zero like the training set ? could it be due to class imbalance because the tumor size is small compared to the background ?
enter image description here