TAIL: A deep learning framework enhancing fabric hand evaluation through multimodal data fusion

Abstract

Fabric hand, or the tactile properties of fabric, is a vital consideration in textile quality assessment. Traditional evaluation methods, including sensory tests and instrumental assessments, reflect either human perception or objective data, but each has its own limitations. This study introduces Textile Attribute Integration and Learning (TAIL), a deep learning framework that combines numerical data, images, and videos of fabric to evaluate fabric hand comprehensively. TAIL enhances hand scoring prediction, classification, and decision-making by leveraging multimodal data fusion, capturing both human expertise and AI-driven features. Addressing the need for remote fabric evaluation in international trade and virtual environments, TAIL outperforms traditional methods, providing a robust tool for the textile industry. Experimental results confirm its superior accuracy and reliability in fabric quality assessment and decision-making.