Traditionally, computer vision solutions for detecting elements of interest (e.g., defects) are based on strict context-sensitive implementations to address contained problems with a set of well-defined conditions. On the other hand, several machine learning approaches have proven their generalization capacity, not only to improve classification continuously, but also to learn from new examples, based on a fundamental aspect: the separation of data from the algorithmic setup. The findings regarding backward-propagation and the progresses built upon graphical cards technologies boost the advances in machine learning towards a subfield known as deep learning that is becoming very popular among many industrial areas, due to its even greater robustness and flexibility to map and deal knowledge that is typically handled by humans, with, also, incredible scalability proneness. Fabric defect detection is one of the manual processes that has been progressively automatized resorting to the aforementioned approaches, as it is an essential process for quality control. The goal is manifold: reduce human error, fatigue, ergonomic issues and associated costs, while simultaneously improving the expeditiousness and preciseness of the involved tasks, with a direct impact on profit. Following such research line with a specific focus in the textile industry, this work aims to constitute a brief review of both defect types and Automated Optical Inspection (AOI) mostly based on machine learning techniques, which have been proving their effectiveness in identifying anomalies within the context of textile material analysis. The inclusion of Convolutional Neural Network (CNN) based on known architectures such as AlexNet or Visual Geometry Group (VGG16) on computerized defect analysis allowed to reach accuracies over 98%. A short discussion is also provided along with an analysis of the current state characterizing this field of intervention, as well as some future challenges.

In textile/leather manufacturing environments, as in many other industrial contexts, quality inspection is an essential activity that is commonly performed by human operators. Error, fatigue, ergonomic issues, and related costs associated to this fashion of carrying out fabric validation are aspects concerning companies’ strategists, whose mission includes to watch over the physical integrity of their employees, while aiming at enhanced quality control methods implementation towards profit maximization. Considering these challenges from a technical/scientific perspective, machine/deep learning approaches have been showing great skills in adapting a wide range of contexts and, in particular, industrial environments, complementing traditional computer vision methods with characteristics such as increased accuracy while dealing with image classification and segmentation problems, capacity for continuous learning from experts input and feedback, flexibility to easily scale training for new contextualization classes – unknown types of occurrences relevant to characterize a given problem –, among other advantages. The goal of crossing deep learning strategies with fabric inspection processes is pursued in this paper. After providing a brief but representative characterization of the targeted industrial context, in which, typically, fabric rolls of raw-material mats must be processed at a relatively low latency, an Automatic Optical Inspection (AOI) system architecture designed for such environments is revisited [1], for contextualization purposes. Afterwards, a set of deep learning-oriented training methods/processes is proposed in combination with neural networks built based on Xception architecture, towards the implementation of one of the components that integrate the aforementioned system, from which is expected the identification of presence/absence of defective textile/leather raw material at a low-latency. Several models powered by Xception were trained with different tunning parameters, resorting to datasets variations that were set up from raw images of leather, following different annotation strategies (meticulous and rough). The model that performed better reached 96% of accuracy.