IARPA and NIST offer $50,000 for best activity detection in extended video

Intelligence Advanced Research Projects Activity (IARPA) is holding an Activities in Extended Video (ActEV) Prize Challenge in collaboration with NIST, offering $50,000 for the technology that proves the best at identifying what people are doing.

The challenge opened in December, and the first stage ends February 28, 2019, with the eventual winner announced in May, and presenting at the Conference on Computer Vision and Pattern Recognition (CVPR) in June.

Participants submit their algorithms for tests of their effectiveness at automatically detecting and localizing 18 different types of activities in extended video footage. The videos include lengthy spans with no activity, and potentially intervals with multiple activities happening at once. The ActEV-PC consists of a qualifying stage, with an open leaderboard evaluation of all eligible solutions, and an independent evaluation stage, with the top eight algorithms on the leaderboard tested on sequestered data by NIST.

The leaderboard currently shows results from 21 entrants.

The ActEV-PC is intended to help develop and evaluate technologies that can help sift through the vast quantities of video data being created and stored, but The Sociable suggests that in government or law enforcement hands, activity recognition could face a backlash from the public, and compares it to facial recognition.

Among many other projects, IARPA is backing research into spoof-proofing biometric authentication systems.

Article Topics

artificial intelligence  |  computer vision  |  IARPA  |  NIST  |  video analytics