PILOT AI LABS, INC. — Department of Defense SBIR Phase I: J201-CSO1

Description

Today’s analysts spend increasing hours exploiting full-motion and still imagery, leaving thousands of hours on the cutting room floor. This leaves significantly important intelligence information out of U.S. hands. If the NGA adopted current commercial Artificial Intelligence (AI) technology, it would allow their FMV and remote sensing imagery analysts to process imagery, in greater quantities, faster, and even at the edge, using low-power computing. Furthermore, this would allow analysts to assess the information and make decisions much faster, bolstering NGA intelligence, surveillance, and reconnaissance (ISR) mission effectiveness. Pilot AI develops computationally efficient neural networks for performing object detection, classification, and tracking within Full Motion Video (FMV) and still imagery. The combination of high accuracy, made possible by neural networks, and computational efficiency has enabled Pilot AI’s software-only solution to be broadly deployed commercially. Facilitated by partnerships with Qualcomm, ARM, and NXP, Pilot AI’s proven commercial solution is already delivering enhanced capabilities on millions of consumer and commercial devices around the world. Pilot AI’s core technology is threefold. The first aspect is state-of-the-art deep learning-based computer vision. Second is the accelerated model building blocks that have a unique adaptive neural network structure that dynamically scales with the complexity of input. Finally, Pilot AI’s software is hardware agnostic and runs on all existing commodity chipsets ARM, x86, NVIDIA, DSPs, Neural Net ASICs, and FPGAs.