Automatic terahertz recognition of hidden defects in layered polymer composites based on a deep residual network with transfer learning

We demonstrate a deep residual network with transfer learning strategy for automatic terahertz (THz) recognition of the hidden defects in fiber reinforced polymer (FRP) composites with small-scale training data. The recognition performance with high accuracy, precision, sensitivity and specificity indicate the effectiveness of the proposed method for automatically identifying different defects in THz nondestructive applications.