A Deep Ensemble Classifier for Surface Defect Detection in Aircraft Visual Inspection

What is it about?

Classification errors made by deep learning algorithms used in automated defect detection can be reduced by combining the predictions of multiple algorithms through ensembling. Multiple stacked ensembles of convolutional neural networks are developed to detect manufacturing surface defects. The results shows that ensembles can improve surface defect detection accuracy through the selection of error-diverse component models.

Featured Image