NASIC GXK STRATFI Proposal - OCR enabled intelligence
AFWERX · AFWERX STRATFI · AFWERX
Award
Description
Vannevar Labs’ Decrypt platform with Optical Character Recognition (OCR) technology enables users to extract foreign language information from documents and images and exploit the data in English. Under this effort Vannevar will extend our OCR capability to support Chinese language documents and optimized for mission relevant technical information. The system will perform text detection then text transcription using deep learning architectures like Convolutional Neural Networks (CNN) and Bidirectional Recurrent Neural Networks (RNN). The models will be trained and deployed on commercially developed GPU processors, providing for fast and accurate transcription and translation. We collaborated with NASIC to test and evaluate our technology’s performance against relevant unclassified and classified data for this mission set. Together, we identified the specific enhancements necessary through this STRATFI project to enable NASIC and its partner organizations to integrate our technology into their classified analytics pipeline. Decrypt has already delivered access to thousands of foreign language end points for NASIC users to exploit through Phase II. As an expansion of the Phase II, the platform will support both Russian and Chinese new sources and download pdf documents of translated Russian and Chinese technical information. NASIC intends to use additional funds from the STRATFI to integrate Decrypt for use on the high-side integration environment. At an enterprise level, NASIC is interested in funding a Phase III contract after development is complete to pursue authority to operate (ATO) on their production software systems and integrate our software into their pipeline through Application Programming Interfaces (APIs). They intend this eventual Phase III to be an IDIQ contract type to enable other Air Force partners, including 16th Air Force, that face similar foreign data exploitation challenges to choose the level of service they need from the technology. Vannevar Labs was selected for a direct-to-Phase II SBIR in early 2021 and submitted a design document in April 2021 – AI-Driven Processing of Foreign Adversary Weapons and Threat Information for Air Force Mission Planning – which listed seven key design requirements. These requirements focused on two primary areas: 1) consistent access to Vannevar’s Decrypt platform, including using the interface to view the baseline Russian Optical Character Recognition (OCR) model test documents; and 2) using NASIC-Vannevar partnership and development expectations to meet the enhanced Russian OCR model milestones.