BMEII-AI / RadImageNet

RadImageNet, a pre-trained convolutional neural networks trained solely from medical imaging to be used as the basis of transfer learning for medical imaging applications.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

mixing of training/validation weights

wheatpennymaster opened this issue · comments

The model weights are not re-instantiated before each call of run_model() in the folds loop. Therefore, the model has seen images from fold 0 (i.e. training set of fold 0) which are included in the validation set for fold 1. This results in an increase in validation performance after each fold is trained.

Heads up, the training code for sarscovid2, pneuomina, hemorrhage, covid19 are also still bugged. Thanks for correcting the others.