The use of the deep learning approach in the textile industry for the purpose of defect detection has become an increasing trend in the past 20 years. The majority of publications have investigated a specific problem in this field. Furthermore, many of published reviews or survey articles preferred to investigate papers from a more general perspective. Compared with published review publications, this study is the first up-to-date study that investigates the implementation of deep learning approaches for the detection of fabric defects from 2003 to the present. As the main objective of this study is to review deep learning-based fabric defect detection, the publications regarding fabric defect detection by using deep learning are examined. The methods, database, performance rates, comparisons, and architecture type of these works were compared with each other. The most widely used deep learning architectures customized deep convolutional neural networks, long short-term memory, generative adversarial networks, and autoencoders. Besides the use of the most used deep learning algorithms, the advantages and disadvantages of these approaches have also been expressed.

Fabric quality has an important role in the textile sector. Fabric defect, which is a highly important factor that influences the fabric quality, has become a concept that researchers are trying to minimize. Due to the limited capacity of human resources, human-based defect detection results in low performance and significant loss of time. To overcome human-based limited capacity, computer vision-based methods have emerged. Thanks to new additions to these methods over time, fabric defect detection methods have begun to show almost one hundred percent performance. Convolutional Neural Networks (CNNs) play a leading role in this high-performance success. However, Convolutional Neural Networks cause information loss in the pooling process. Capsule Networks is a useful technique for minimizing information loss. This paper proposes Capsule Networks, a new generation method that represents an alternative to Convolutional Neural Networks for deep learning tasks. TILDA dataset as source data for training and testing phases are employed. The model is trained for 100, 200, and 270 epoch times. Model performance is evaluated based on accuracy, recall, and precision performance metrics. Compared to mainstream deep learning algorithms, this method offers improved performance in terms of accuracy. This method has been performed under different circumstances and has achieved a performance value of 98.7%. The main contributions of this study are to use Capsule Networks in the fabric defect detection domain and to obtain a significant performance result.

Recognition of emotions by analyzing the facial expressions has been an important research area in last decade. Existing studies in the literature mostly seek for a classification model that is linking the emotions with the facial expressions that are defined by the basic components of face (such as eyes, mouth, and ear). In this regard, Artificial Neural Networks (ANN) has been a widely employed approach for classification purposes. Being different than the existing approaches, we searched for the points/landmarks that can contribute to accuracy of ANN model. With this purpose we have started our search with the 153 possible distances among 18 critical candidate points/landmarks. The 16 of these distances having significant contribution to the model accuracy was selected by using the correlation-based feature subset selection (CFS) method. The results indicated that our ANN model has 91.2% correct classification rate.