Design and Development of Handloom Saree Testing System using Machine Learning

Silk sarees have been a part of Indian culture for centuries and are considered as traditional attire of entire women folk in India, as every weaving cluster over 200 has its uniqueness. Silk sarees are known for their exude elegance, lustrous beauty, sophistication, luxurious feel and durability. Handloom and powerloom are the two possible ways of weaving a silk saree. Handloom sarees are softer and more resilient compared to power loom sarees, whereas powerloom sarees are stiff and do not drape well. Because of these qualities, handloom sarees are preferred over powerloom sarees, and are hence expensive. In this paper, a novel attempt is made to design a silk saree testing system using an unsupervised machine learning algorithm called k-means clustering algorithm and a supervised machine learning algorithm called Decision Tree classifier. The proposed testing system is capable of detecting whether the saree under test is woven using handloom or powerloom. Two Decision Tree classifier models are designed with their performances compared and evaluated by testing the proposed system at various silk saree stores in and around Bangalore. The proposed system will serve as a reliable objective testing method over conventional inefficient subjective testing methods, giving scope for the popularization of proposed system through authentic textile testing laboratories in India and its recognition from the Bureau of Indian Standards (BIS) as a validated testing method. The proposed system is an outcome of a funded project with Central Silk Board under Ministry of Textiles, India. The proposed system yielded a maximum recall value of 0.92 with a response time of 81 seconds and costs only $100.