Component Detection and Evaluation Framework (CDEF): A Semantic Annotation Tool

Abstract

AbstractImage labeling is the process of manually assigning a class to subregions within an image for machine learning applications. When these subregions are complex shapes, this process is known as semantic segmentation. We propose a new software application, the Component Detection and Evaluation Framework (CDEF), for creating such semantic labels. The benefits of CDEF over existing tools are highlighted, and further improvements are proposed.