ARETE ASSOCIATES — Department of Defense SBIR Phase I: AF182-006

Description

Areté Associates proposes use recent advances in deep learning to streamline the processing, exploitation, and dissemination (PED) of the intelligence with an automated algorithm capable of achieving positive identification and characterization of objects of interest. Such an algorithm will significantly reduce the workload of the analysts. Areté will leverage our Attribute Localization and Instance Extraction Neural (ALIEN) network to simultaneously detect, localize and characterize data, thus augmenting the utility of a human analyst. In addition to increasing detection accuracy and precision, this capability exploits technological advances in Tensorflow protobuf processing architectures to be able to retrain and redeploy the neural network without affecting the operational product, continuously improving the performance and further reducing the false alarm rate. During the course of Phase I, Areté will participate in the technology accelerator program and will conduct a feasibility study to determine the potential utility of the ALIEN network on a relevant problem set within the Air Force’s ISR mission area. During Phase II, Areté will further develop and demonstrate a prototype of the ALIEN network to enhance the ISR mission with validated utility.