search

spockbuddha / OOAMS

on-orbit servicing, assembly, and manufacturing (OOAMS). Management of costly space-based assets like telecommunications satellites in geosynchronous orbit, mitigation of orbital debris and lifting of more payload to orbit are among the drivers. Evolving past the dimensional limits imposed by launch vehicle fairings for low mass-to-volume ratio payloads motivates the remaining OOAMS segments: on-orbit assembly and on-orbit manufacturing. Utilizing on-orbit manufacturing to achieve size, weight, and power reductions through the integration of multiple functions like thermal and electrical systems into structures or through structural optimizations
0 stars 0 forks source link

ARL (CRA) - Distributed and Collaborative Intelligent Systems Technology Collaborative Research Alliance • Project Maven #4

Open spockbuddha opened 5 years ago

spockbuddha commented 5 years ago

–Phase 1: Develop and integrate computer vision algorithms needed to augment and assist military and civilian analysts with high volume Full-Motion Video (FMV) data primarily from Unmanned Aerial Vehicles (UAVs); program delivered first algorithms to military systems in December 2017 –Phase 2: To partner with industry and academia, on October 24, 2017 the AWCFT led by the Under Secretary of Defense for Intelligence (USD(I)), partnered with the Army Research Laboratory to host the first Project Maven industry day • Intent is to expand work beyond FMV to all areas of actionable intelligence • Air Force Lt. Gen. John N.T. “Jack” Shanahan, director for Defense Intelligence for Warfighter Support at USD(I), stressed the need for DOD to partner with industry, academia, and national laboratories as it goes about “operationalizing AI and machine learning for the warfighter.” • Lt. Gen Shanahan proposed the need for a DOD “Center for AI” to serve as a clearinghouse for data and opportunities, and expressed a desire for a consortium of people to help DOD understand potential new capabilities. The team has visited laboratories and top universities to learn about computer vision and how DOD could better employ it in its data processing, exploitation, and dissemination (PED) enterprise

spockbuddha commented 5 years ago

• Computer Vision models that enable Geospatial Intelligence processing and exploitation in constrained and unconstrained compute environments through: object identification, object classification, object localization, unique object recognition/recall, object pixel georegistration, object tracking, semantic segmentation, logical expression or semantic description, and activity/situation recognition. • New data labeling techniques, tools, and tradecraft for data annotation in support of deep learning: "edge" or "labeling on the line", use of synthetic or photorealistic data, and relabeling/retraining in near-real time. • Interfaces for the display, search, and interaction with algorithmically derived metadata and tabular structured algorithmic output: anomaly and pattern of life analysis, object search (visual and metadata), visualization, and fusion with other structured data. • Storage and indexing capabilities for local algorithmically-produced data. • Language algorithms to process verbal form and written text, including, but not limited to: natural language processing, automated language translation, and sentiment detection.