Institute of Information Science, Academia Sinica

 Computer vision has been widely used in discovering patterns from complex and unstructured data such as videos and images. Successful techniques need vast amounts of data and labels for training and validation. Creating datasets and labels require significant efforts.  A team at Purdue University creates datasets using global public cameras that can provide real-time visual data. These cameras can continuously stream live views of national parks, zoos, city halls, streets, university campuses, highways, shopping malls, and so on.  The stationary cameras have contextual information (such as time and location) about the visual data. By cross-referencing with other sources of data (such as weather and event calendar), it is possible to label the data automatically. This system is a foundation for many research topics related to analyzing visual data, such as (1) how can this system automatically produce labels for computer vision, (2) how to automatically place cameras to meet restrictions of computer vision, and (3) how to protect privacy of video streams (real-time visual data) used in computer vision.