SkyShot: target detection through spatiotemporal context

Satellite platforms play a critical role in the modern defense and intelligence infrastructure, providing timely, detailed, and readily available imagery to support U.S. national security. A fundamental requirement for this type of data is automated target detection and recognition, helping analysts to quickly locate and correctly identify targets of interest from vast quantities of image data. Despite significant progress in object discovery based on modern deep learning techniques, target detection algorithms designed for one image source often fail to generalize to more challenging conditions.  VSI proposes SKYSHOT, an overhead image analysis capability designed to improve target detection and characterization through the innovative use of spatiotemporal context. This effort will realize significant tactical and strategic advantage for the U.S. Government, promising to reduce workloads and improve intelligence yields for analysts monitoring satellite imagery to produce strategic analysis and situational awareness assessments.  SKYSHOT seeks to develop an overhead image analysis capability to improve target detection and characterization in a wide range of diverse and degraded satellite imagery, leveraging rich spatiotemporal scene context available from image metadata, geospatial information, and 3D scene geometry. SKYSHOT is centered around annotation transfer, where target locations discovered in high-quality imagery are rapidly projected to degraded.  VSI is uniquely suited to this project, offering expertise in remote sensing, object detection, 3D reconstruction, and machine learning; extensive experience in the development of innovative software solutions for challenging defense and intelligence problems; and a thorough understanding of the needs of image analysts across the defense and intelligence community.