T. Emerick, A. Johnson
Commonwealth Computer Research, United States
Keywords: FMV, AI, Automated Captioning, Compression, Tactical Video ProcessingRVLADE represents a State of the Art set of video processing algorithms that employ deep learning models, co-located with tactical collection system to pre-process video for transmission. Using GPU enabled “system on a chip” (SoC) platforms, RVLADE's on-platform processing will provide better cues to users while simultaneously reducing bandwidth usage to the control system, enabling the distribution of an innovative video analysis algorithm to the tactical edge. In default mode and in near-real time RVLADE will transmit a low-bandwidth text description of the scene as it takes place in the video. Upon request, users are able to connect to the video stream, beginning at some point of their own selection in the near past. As a result, RVLADE reduces the average bandwidth needed for video transmission and reduce analyst time spent studying irrelevant video footage. As a result, the efficiencies obtained will enable the deployment of additional video sensors if so desired. Further, RVLADE’s ability to generate a mission narrative will be a game-changing improvement over current alarm-based systems. RVLADE’s video-to-text technology is based on CCRi’s VLADE system which has demonstrated this capability in a Phase II Air Force Research Lab SBIR.