Vision-Based Carpet Similarity Inspection Using Deep Learning And Genetic Algorithms

An automatic carpet similarity inspection system based on computer vision and machine learning is developed in this article to replace the traditional inspection approach using human eyes in the carpet industry. First, each carpet image is extracted a feature vector using a popular deep learning model (Inception-V3). Then the unsupervised clustering learning algorithm is employed to automatically assign the carpets to different groups based on their similarity. Finally, a genetic algorithm-based approach is proposed to find the optimal arrangement of the carpets in each group, as required by the industrial carpet plant. The experimental results validate that the proposed approach is successful and effective.