Deep Learning for Fashion Automation
Deep learning networks are known for its ability for their highly accurate image recognition results, but they need large dataset to function at commercially viable accuracies with acceptable predictability. In fashion technology where apparel designs are short-lived and the definition of individual attributes is very ambiguous, the challenge lies in designing a Deep learning system which can learn various subtle attributes about an apparel with insignificant amount of data per class and give respectable accuracies for commercial use. Hence a distributed problem solving approach is taken with multiple CNN’s networks working in sync to provide desirable results with less training size and predictable results.
Vipul Divyanshu heads R&D at Voonik Technologies, India's 2nd largest fashion marketplace. He previously co-founded Trialkart, a mobile virtual-trial and AI driven technology startup, with Jayalakshmi Manohar and Harsha M, which was acquired by Voonik in the 6th month of its operation. At Voonik, he integrated TrialKart’s technology and further innovated to develop a state-of-art Deep learning stack, which automates all image related tasks across all verticals of the company. His notable feat includes a tailored system that fits the need of high speed Image tagging, recognition and search across millions of products. He is a lifelong learner and a strong AI believer and hopes to revolutionize the world with AI.