SEMA: Segmentation and Estimation of Missile Attitude
Project runs from 08/02/2021 to 01/31/2022
To develop a technology called “Segmentation and Estimation of Missile Attitude (SEMA)”. The solution is based on a highly effective hybrid architecture that combines top-down knowledge of 3D object CAD models with deep learning-based bottom-up object signature inference, which makes the approach robust to resolution, occlusion, changes in viewpoint, and other covariates. The key innovation of the proposed approach is to formulate the object segmentation problems in a hierarchical learning-based framework which uses the 3D structure of the object.