Video represents a powerful tool for enabling high quality debriefs or after action reviews (AARs), however there are many challenges to utilizing video for debriefing. One key challenge is the necessary effort to find the specific time (bookmarks) in which important activities occur (i.e.; highlights). To alleviate these issues, we have developed a system for recording, ingesting, and automatically bookmarking video from person-worn cameras during military medical training exercises. By utilizing person-worn cameras, we are able to eliminate the requirement of having a separate videographer. This system was used at Joint Base Lewis-McChord to record exercises as part of a recent study. Our system was able to document all of the events that took place without any loss of data. Furthermore, using the algorithms we developed the system was able to bookmark key medical activities such as when tourniquet application, gunshot wound treatment, and needle chest decompression where performed with an accuracy of X% (97% for tourniquets, Y% for chest decompression, and Z% for gunshot wound treatment; some of these numbers are still being computed and will be ready for the paper review but aren’t currently ready). This video was later presented to instructors where it provided sufficient detail to accurately assess performance. On average, the video and associated bookmarks were available for review within X:XX minutes of turning in the camera, even though the average length of the video was XX:XX minutes (a processing speed that is X% of real-time; of these numbers are still being computed and will be ready for the paper review but aren’t currently ready). In addition, advanced features of the system are able to provide assessment metrics, such as estimated time to complete a tourniquet application.
Keywords
AUTOMATION,COMBAT CASUALTY CARE,DEEP LEARNING,EDUCATION
Additional Keywords