CLOSTRA INC — Department of Defense SBIR Phase I: CBD222-005

Description

Clostra’s 3DGEN is an end-to-end process for automated reconstruction of high-fidelity three-dimensional (3D) models from two-dimensional (2D) images. It leverages proven machine learning techniques for fusion of input images and generation of 3D models and fills a current technological gap in Warfighter training and operation in the domain of Chemical and Biological Defense. The outcomes of 3DGEN include the generation of 3D models of objects, buildings, and terrain via fusion of 2D images, reconstruction of topologically accurate 3D models when blueprints are available, and industry standard outputs that can be integrated with relevant simulation tools. The success of the project will enable Warfighters to train and gain situational awareness during and before operations. 3DGEN will be built upon foundational work performed for Clostra’s DEEP Screen and DeepMeet projects, in which 2D images/videos are successfully generated into 3D virtual reality environments. In Phase I, Clostra will merge two 2D image types (visible light camera images and blueprints) and use the 2D images to generate 3D models of objects, buildings, and/or terrain. We will generate parameter files that can be used for CONTAM or other simulation software. A proof-of-concept and development pipeline for an inferenced building generation model will be created.