A Deep Learning Solution for Optical Fragment Mass Estimation Using Holistic Data

Protection Engineering Consultants (PEC), in collaboration with Southwest Research Institute (SwRI) and Talon Analytics, proposes to develop the DeepMass Utility and Plugin—a validated, field-ready software solution for estimating fragment mass from 3D stereoscopic high-speed video (HSV) data. DeepMass will replace time-consuming post-test methods with an automated, deep learning-based approach that delivers rapid, calibrated mass predictions directly from imaging diagnostics. The Phase I effort successfully demonstrated a proof-of-concept DeepMass-X model trained on synthetic data, achieving superior performance compared to conventional mass estimation methods. In Phase II, PEC will build on this foundation by integrating high-fidelity training data from physical, synthetic, and stereoscopic sources, enhancing the model architecture to address small-fragment limitations, and expanding capabilities with fragment classification, uncertainty quantification, and computational optimization.A core deliverable will be the DeepMass Utility—a standalone, user-friendly application enabling engineers and analysts to generate, train, and evaluate models with minimal deep learning expertise. In parallel, PEC will develop the DeepMass Plugin for integration into Navy fragment tracking systems, streamlining use within existing workflows. A modular RPC microservice architecture will support scalable deployment, including cross-platform CLI and GUI interfaces. Transition and commercialization will be pursued through direct engagement with Navy test ranges (e.g., NAWCWD, NSWC Dahlgren, Crane, Eglin AFB), where DeepMass will support weapon effects diagnostics, lethality modeling, and test planning. A structured Technical and Business Assistance (TABA) partnership with Dawnbreaker, Inc. will guide market validation and customer discovery across both defense and commercial sectors. Targeted commercial markets include aerospace, energy, and industrial safety organizations conducting explosive and high-speed diagnostics. By the end of Phase II, PEC will deliver a validated DeepMass software suite, integration-ready plugin, trained model packages, and full user documentation. These capabilities will enable operational deployment, facilitate transition through Phase III contracts, and support long-term adoption within DoD and industry. DeepMass will ultimately improve test throughput, reduce analysis timelines, and increase the accuracy and confidence of fragment mass data used in warhead characterization and risk assessment.