BENEFITS OF ADOPTING MACHINE LEARNING
Today every industry is moving faster. Facing the booming Fast Fashion, the textile industry in Taiwan has to keep up the pace, considering how to introduce products in a timely manner so as to adapt to the new trend. However, speaking of the operating mode of the traditional textile industry, it often takes a great deal of labor and time to go through different steps such as designing, choosing and picking, spinning, weaving, cloth finishing, and so on. Specifically, it can take 40 to 45 days to manually differentiate various fabrics. Should it be any errors in picking or differentiating process, one could imagine spending another three months reproducing before the finished products could be handed over to the customers. Additionally, people believe that traditional approaches of warehouse management are incapable of addressing thousands of textile samples. This not only makes it difficult to preserve the previous producing experience and techniques, but it’s also time-consuming to search for samples and to manage stocks. The techniques and samples thus would be hard to re-utilize.
Our partner－Hermin Textile aspires to combine technological intelligence with specialized experience and techniques accumulated over years and to properly preserve them via systematic data. This benefits future analysis and management; it also helps enhance operating efficiency. Thus since last October, Hermin Textile has begun to execute digitalization and filing of previous fabric samples using TensorFlow and Google Cloud Platform. Furthermore, Hermin adopted machine learning technology to train and structure a recognition model, working with designers to swiftly search and locate specific patterns in its database.
We are delighted to see Hermin Textile leverages Google Cloud platform and machine learning to effectively simplify its old operating methods and process. While it used to take 1.5 – 3 months to go through initial inspiration, designing, sample choosing, and sample offering, now it only takes 2 – 3 days to complete these steps. In the meantime, it is estimated that the time needed to introduce a new design will be considerably reduced by 25％, from the previous 12 months to now only 9 months.
Hermin Textile is also planning on a sample-collecting app which would be built on machine learning and Cloud services; coupled with social networking functions, the app aims to modernize the communication between designers, reutilize past manufacturing expertise, and to facilitate the overall circular economy of manufacturing including the textile industry.
The case of Hermin Textile is arguably the pioneer of traditional industry’s transformation through technology intelligence in Taiwan. Google Cloud will keep improving the existing Cloud machine learning services, expecting to gradually expand to other industries and to help make enterprises in Taiwan even more competitive.