IARPA posts new BAA for DIVA program
On September 13, the Intelligence Advanced Research Project Activity posted a new broad agency announcement for the Deep Intermodal Video Analytics (DIVA) program (IARPA-BAA-16-13). The proposal due date for the initial round of selections is 5:00pm Eastern Standard Time on November 7, 2016.
The volume of video data collected from ground-based video cameras has grown dramatically in recent years. For mounted cameras with rigid motion, such as pan-tilt-zoom, the video data is predominantly collected for security, public safety, transportation and infrastructure monitoring, and primarily used for forensic, legal and insurance purposes.
The need to guard against theft, crime or attacks continues to rise. However, there has not been a commensurate increase in the usage of intelligent analytics for real-time alerting or triaging video. In many cases, security personnel or operators of camera networks are overwhelmed with the volume of video they must monitor, and cannot afford to view or analyze even a small fraction of their video footage. Yet, the task of monitoring video in airports, at border crossings, or at government facilities becomes increasingly critical each year.
In addition, when incidents do occur and officials are tasked with forensically analyzing large volumes of video, it is manually intensive, to identify relevant activities and the subjects of those activities. DIVA aims to develop technology to automate much of this analysis.
While there has been recent advancement in the development of image recognition techniques, this advancement has been predominantly limited to applications outside the security video realm. First, the focus has been on image and video from social media, and in special commercial applications such as autonomous vehicle navigation. Second, there has been little focus on understanding activities in video that depend on temporal reasoning across video frames. Third, few analytics detect activities in a multi-camera streaming environment and have the ability to alert in real-time.
DIVA will significantly push the state-of-the-art in three areas:
Automatic detection of activities, as well as persons and objects, in cluttered scenes,
Temporal reasoning of video to greatly improve activity detection,
Activity detection and scene understanding from overlapping and non-overlapping camera viewpoints.
Full information is available here.