“…These machine learning approaches classify wood defects by factoring the statistical variations of the defect images to learn about the desired defects with the assistance of several classifiers such as neural networks [59], k-nearest neighbors (k-NN), decision trees and SVM [17]. On the contrary, deep learning has been shown to be highly effective in a wide range of image-based applications, including object detection and identification, facial detection and pattern identification due to their network flexibility in discovering custom defects based on the dataset [60]- [64]. Furthermore, feature extraction for deep learning is embedded in the learning algorithm, where features are extracted in a fully-automated manner, without requiring any intervention from a human expert.…”