Gender classification of full body images based on the convolutional neural network

Gender classification is one of the most interesting and challenging problems in computer vision and has been widely studied based on facial images. However, the images of human we taken from the real-world surveillance are mostly full body and relatively blurry, which is much more difficult to classify due to different poses and backgrounds in unconstrained scenarios. In this paper, we propose a new network structure based on a convolutional neural network (CNN), which is less complicated and has a small number of layers. Moreover, it can achieve a high accuracy with even trained with limited data. We evaluate our method on the dataset collected from real-world video surveillance and compare various learning algorithms including Alex Net and Google Net. The experimental results showed that the proposed model achieved better results than the tested state-of-the-art network structures.