Fig. 1
From: Quantifying dislocation-type defects in post irradiation examination via transfer learning
An overview of the transfer learning training process for dislocation defect quantification is summarized here. Two transfer learning models are evaluated in this work and both begin with pre-trained MSCOCO weights. Using the pre-trained weights, model A is trained on a dataset consisting only of MA956 ODS micrographs. In contrast, model B is trained on the crops26 and synthetic cavity27 datasets. The best weights from this intermediate stage are then used to train on MA956 ODS training images which produce the weights of model B.
