An Experimental Evaluation of Sparse Classifiers for Human Detection

The paper presented an experimental study on human detection in image. The experiments are attempted in regard of its sparseness. We consider two contexts of sparseness: classifier design and feature (components) selection. The results show the Bayesian treatment might conduct sparsity on classifier in these two differenct contexts. The analysis is performed by assessing the DET-curve to get insight on the trade-off between the effort for reducing miss-rate while reducing false-positive rate. D. Aurrahman, N. A. Setiawan, C. Oh, C. Lee, An Experimental Evaluation of Sparse Classifiers for Human Detection, 14th Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV 2008), pp. 148-153, Jan. 2008