Ground-based cameras acquire image and video data of targets in various stages of flight. Cluttered backgrounds, poor contrast, low resolution and other imaging challenges make it difficult to discern the target and estimate its attitude (pose). In this effort we propose a deep learning based approach to detect and segment targets of interest. The results of segmentation will be visualized as a pixel-wise mask. EO/IR imagery and simulated data will be used to demonstrate feasibility of the approach.