Visual attribute classification using feature selection and convolutional neural network
Rongqiang Qian, Yong Gang Yue, Frans Coenen, Bailing Zhang · 2016
Visual attribute classification has been widely discussed due to its impact on lots of applications, such as face recognition, action recognition and scene representation. Recently, Convolutional Neural Networks (CNNs) have demonstrated promising performance in image recognition, object detection and many other computer vision areas. Such networks are able to automatically learn a hierarchy of discriminate features that richly describe image content. However, dimensions of features of CNNs are usually very large. In this paper, we propose a visual attribute classification system based on feature selection and CNNs. Extensive experiments have been conducted using the Berkeley Attributes of People dataset. The best overall mean average precision (mAP) is about 89.2%.