情報処理学会 第80回全国大会講演要旨

To design a good human detector, we may collect a huge number of data, and train the detector off-line. However, even if the training data set is very large, it may not contain enough information for some specific environment, and the obtained model may not generalize well. In this paper, we study incremental learning of a support vector machine-based human detector in an office environment. Experimental results show that it is possible to obtain a good human detector customized to a certain environment with less data if we use incremental learning.