An Automated Model for Fabric Weave Pattern Recognition and Classification Using Deep Learning Based Image Processing

Fabric weave type identification has many uses that adds to importance of advance classification methods; e.g fabric weave should be known in order to identify the most suitable use. But the manual inspection of fabric causes human errors such as, stressed, less focused, eye focus and mishandling of the apparatus. There are lighting effects which also degrades the performance of manual inspection. In recent years, computer vision based approaches have achieved remarkable performance in the fields of object recognition and image processing, thus opening a new door to develop an automated system for the recognition of woven fabric patterns for production of fine quality fabric according to customer needs. In this paper, we have devised an automatic weave pattern recognition mechanism based on convolutional neural networks(CNN). Initially three basic weave patterns were chosen i.e Twill, Plain and Satin. The results were analysed for its possible adaptability in the task of fabric weave pattern recognition. The classification results give the average accuracy of 72^% for all three samples whereas individually, Plain weave gives 100% accuracy, Twill gives 66% accuracy and satin weave gives 50% only, which indicates a room for improvement in future.

I. Introduction

Fabric industry has been one of the oldest in human history [1]. The need is never ending and technologies relating to textile has ever been growing since the beginning of modern era. With advancements in technology there has been remarkable progress in Textile industry, however the demand for improvement has become manifolds as well.