Human face detection adopting small scale convolutional netural network module in embedded system

The invention provides embodiments of Human face detection adopting a small scale convolutional netural network (CNN) module in embedded system. The module is configured into a multi-task cascading CNN module. In some embodiments, the CNN system based on a sub image can be configured to be large scale CNNs requiring no division and processing whole input images, so that output of the CNN system based on the sub image can be completely identical to that of large scale CNN. Therefore, the CNN system based on the sub image and technology are applied to one or more stages of a cascading CNN or a multi-task cascading CNN (MTCNN), so that larger input images input into the specific stage of the cascading CNN or the MTCNN can be divided into a group of sub images having smaller sizes. Therefore, each stage of the cascading CNN or the MTCNN can adopt identical small scale hardware CNN modules which are under maximal input image size limitations. 本申请实施例提供了基于使用小规模硬件卷积神经网络(CNN)模块的人脸检测系统的多个示例，其中该模块被配置到多任务级联CNN模块中。在些实施例中，