SBIR / STTR #A214-047

Army SBIR AI/ML solution for automated height-of-burst video scoring

The U.S. Army SBIR opportunity seeks an AI/ML computer-vision approach to automate height-of-burst scoring from test video. - Government buyer: U.S. Army SBIR; no specific agency office or sub-agency is identified. - Requested effort: Develop and demonstrate algorithms that automatically estimate height of burst from standard video files, reducing manual video review and enabling faster, potentially real-time scoring. - Train and evaluate the algorithms using existing verified datasets and test them across multiple experiments and projectile use cases. - Support field deployment for proximity-fuze and munition testing, measuring changes in time, cost, and performance. - OEMs and vendors: No procurement OEMs or vendors are identified. An attached research paper is attributed to New York University’s Multimedia and Visual Computing Lab; it describes distance-estimation research, not a named supplier for this opportunity. - Attachment models: Base Model (res50), Base Model (vgg16), Enhanced Model (res50), and Enhanced Model (vgg16). The attachment gives no quantities or part numbers. These models estimate object distance from monocular RGB images and are not stated as required procurement items. - Notable technical context: The attached research uses extended KITTI and nuScenes(mini) datasets and reports improved distance estimation from an enhanced model using a keypoint regressor and projection loss. The solicitation does not specify that these datasets or models must be used. - Performance period and funding: Up to 18 months; funding is stated as up to $1.7 million.

Description

The solicitation seeks an AI/ML-based computer vision approach to automatically score height-of-burst testing for fuses using standard video files, reducing the manual effort required to review and measure test footage. The work includes identifying height-of-burst requirements for projectile use cases, training algorithms on existing verified datasets, and demonstrating performance across a wide range of experiments. Direct to Phase II requires demonstrating relevant algorithms, data, and reports; Phase III includes field-testing deployment and quantifying time, cost, and performance outcomes. The topic lists a duration of up to 18 months and funding of up to $1.7 million.

View original listing