Generative adversarial network with object detector discriminator for enhanced defect detection on ultrasonic B-scans

Non-destructive testing is a set of techniques for defect detection in materials. While the imaging are manifold, ultrasonic one used most. The analysis mainly performed by human inspectors manually analyzing recorded images. low number defects real inspections and legal issues considering data from such make it difficult to obtain proper results automatic image (B-scan) analysis. In this paper, we present novel deep learning Generative Adversarial Network model generating B-scans with distinct locations. Furthermore, show that generated can be synthetic augmentation, improve performance convolutional neural object networks. Our method demonstrated on dataset almost 4000 more than 6000 annotated defects. Defect when training yielded average precision 71%. By only increased 72.1%, mixing achieve 75.7% precision. We believe generation generalize other challenges limited datasets could personnel.