Nikita Manovich, Senior Software Engineer at Intel, presents the “Data Annotation at Scale: Pitfalls and Solutions” tutorial at the May 2019 Embedded Vision Summit.
In many real-world use cases, deep learning algorithms work well if you have enough high-quality data to train them. Obtaining that data is a critical limiting factor in the development of effective artificial intelligence.
In this talk, Manovich identifies common pitfalls encountered in obtaining and using public and private data for training and evaluating deep neural networks for visual AI—and presents techniques to overcome these pitfalls. He also presents the open source Computer Vision Annotation Tool (CVAT) (github.com/opencv/cvat), illustrating techniques his company has implemented to streamline annotation of visual data at scale. He discusses challenges faced in developing CVAT, how they were addressed, and plans for further improvements.
