Learning Object and Scene Recognition Strategies

摘要

This report summarizes activities performed as follows: (1) development of theSchema Learning System (SLS) from Draper thesis; (2) learning multi-variate decision trees for object classification; (3) real-time color classification in U6V domain via table look-up; (4) interactive real-time classification visualization for training data specification; (5) new formulation of approach to learning via Markov decision processes.