A Case Study on Master Students with the Textile Background in Tackling Fiber Identification Problems

摘要

The identification of Wool/Cashmere fibers is an extremely challengeable problem since both fibers possess highly similar morphological features. In this research, we monitored two group of master students with the textile background in tackling this specific task. Both groups have no experience on machine vision. Two learning curves, machine learning, and deep learning were randomly distributed to the groups for them to learn. The entire experimental procedure was earned under the same condition, i.e., computing resources, laboratory facilities, working environment, and weekly discussion with an academic advisor. From our observation, higher accuracy has been reached by the group following the path of deep learning. Comparatively, knowledge of image analysis or hand-crafted feature extraction has empowered the students following the path of machine learning with a more fundamental understanding of feature extraction.