Bengaluru-based AI Startup Infilect Technologies won the prestigious Startup Research Award at KDD 2018 in London earlier this week. The research paper titled, How To Extract Fashion Trends From Social Media: A Robust Object Detector With Support For Unsupervised Learning won the award.
SIGKDD, a premier data science conference organised by the Association of Computing Machinery (ACM) showcases cutting-edge research in the new tech areas from top researchers in the world. Infilect was one of the startups from a list of five to have won the award.
Speaking to Analytics India Magazine, Anand Prabhu Subramanian, co-founder and CEO of Infilect, said, “We are very excited to be recognised for our latest work in object detection which forms a core part of our computer vision tech stack. We have surpassed the state-of-the-art by over 17 percent and have very high accuracy, in turn leading to production-ready solutions for our customers.”
The technology comes as a breakthrough as many in the computer vision space starve for labelled data to train huge neural networks. Speaking about this issue, Subramanian said, “We have devised a clever way of using a large amount of untagged data to first train the base architecture of the object detector using unsupervised learning and then fine tune with a small amount of tagged data to achieve very high accuracy. For instance, in one of our production use cases, a tagged dataset as small as 250 images gave us over 90 percent accuracy.”
Infilect Technologies was co-founded by Anand Prabhu Subramanian and Vijay Gabale (both ex-IBM research scientists) in 2015. They specialise in deriving intelligence by parsing large-scale visual content such as photos and videos using proprietary deep learning and AI technology and provide highly scalable automation and analytics solutions for retail (both online and physical retail) and advertising industries.