Performance Evaluation of a Semantic Perception Classifier

Abstract : In order to assess the technology developed by members of the U.S. Army Research Laboratory s Robotics Collaborative Technology Alliance (RCTA), RCTA s Integration and Assessment team conducts experiments to test the limits of developed technologies. We conducted such an assessment in October 2013 to evaluate the performance of a semantic labeling system developed by researchers at Carnegie Mellon University. The system consists of a Basler AC1300gc camera with Fujinon TF28DA8 lens and processing software, which runs on a laptop computer. The apparatus functions on small robotic platforms, such as a K-bot. This report details the design for our data collection and our assessment of the technology by examining precision, recall, and F measure via exploratory data analysis and analysis of variance (ANOVA).