Cascaded face detectors using convolutional neural network (CNN) are widely adopted in real-world applications due to their superior efficiency. However, the performance of those detectors degrades significantly with the presence of challenging faces. In this paper, inspired by the fact that large receptive field provides more contextual information for improved performance, based on the existing cascaded CNN face detector, we propose a context-aware cascaded face detector, which aggregates more information for each candidate window. We empirically evaluated different methods for enhancing the information embedded in the receptive field and provide detailed comparisons. Moreover, we propose a score fusion strategy to combine the confidence scores from different detecting stages in the cascaded network to get a more reliable confidence score. We also propose an improved strategy to perform non-maximum suppression that achieves better efficiency when the number of proposals is huge. Extensive experiments on FDDB and WIDER FACE datasets demonstrate that our real-time detector achieves better performance over other cascaded detectors.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.