Summary
The development of more advanced behaviors and decision-making schemas for autonomous systems has been constrained by the way training data has been labeled with bounding boxes and object classification. Labeled data limited to bounding boxes and object classification is insufficient for the development of the next generation of intelligent autonomous systems. Autonomous platforms across domains require additional information from the platform perception software to inform autonomous behaviors like object avoidance or maintaining standoff distance localized to a point on the object. Such information may include: Weather conditions (e.g., sunny, cloudy, foggy, and rain); Lighting conditions and source orientation; EO/IR sensor characteristics (e.g., type, resolution, look angle, and distortion factors) ; Object environment and sensor domain (i.e., maritime, aerial, terrestrial, or sub-surface); Object dimensions (e.g., length, and height); Object pose, orientation, and bearing; Segmentation or boundaries of key object features; Percentage of object occlusion ; Semantic segmentation of complex objects and scenes.