Neural network circuit and self-circulation multi-stage iteration method thereof

Abstract

The invention provides a neural network circuit and a self-circulation multi-stage iteration method thereof. The neural network circuit comprises an image reduction unit, a sensitive area image data reading unit, a multi-channel selection unit, a configurable neural network circuit unit, a neural network structure configuration unit, a sensitive area coordinate information storage unit and a coordinate restoration unit, wherein firstly the image reduction unit reduces an original image, a sensitive object area in the image is segmented by the configurable neural network circuit unit, the coordinate restoration unit restores the sensitive object area into the original image, and then the sensitive area image data reading unit performs local reading; and then the read data is sent to the configurable neural network circuit unit for classification and recognition. The recognition process is performed by adopting a self-circulation multi-stage iteration mode, and a continuously refined recognition result is obtained through going deep for multiple turns.

Description

technical field [0001] The invention relates to a neural network circuit structure, which enables it to have self-circulation and multi-stage iteration functions. Background technique [0002] Image sensitive area recognition technology is an existing technology. The common area division network algorithm is FCN (see https: / / blog.csdn.net / taigw / article / details / 51401448). FCN first divides the traditional CNN into The fully connected layer in is converted into a convolutional layer one by one. After multiple convolutions (and pooling), the obtained feature map is getting smaller and smaller, and the resolution is getting lower and lower. At this time, the value of each point of the feature map represents the probability of a classification. When the probability of this point When the value is greater than a certain threshold (the threshold can be configured), it is considered that there is a sensitive object at this point, and then the coordinates of the sensitive area where...

AI Technical Summary

Problems solved by technology

However, due to the complexity of the algorithm of the deep learning neural network itself (including the image segmentation algorithm), it will cause a great consumption of hardware circuits and power consumption, especially mobile device terminals are difficult to support huge deep learning networks.

The current terminal equipment can only be used to calculate the recognition processing of smaller images, such as (320x240, 256x256, etc.), when encountering large image recognition, it can only be handed over to the cloud server for calculation or the calculation time is too long, which makes some applications useless. restricted

Images