Computing method based on neural network and computing device thereof

Abstract

The embodiment of the invention provides a computing method based on a neural network. The computing method comprises the steps that an original image of the tensor form is acquired; fixed point computation is performed on the original image of the tensor form; an image thermal diagram is generated based on the output data of fixed point computation; and the original image is marked based on the image thermal diagram. According to the computing method based on the neural network through the fixed point method, fixed point computation is adopted so that computing burden is low, resource occupation is less and thus the requirement for hardware is low.

Description

technical field [0001] The present invention relates to the field of image processing, and more particularly relates to a calculation method and device based on a neural network. Background technique [0002] Neural Networks (NNs), also known as Artificial Neural Networks (ANNs) or Connection Model, is an algorithmic mathematical model that imitates the behavioral characteristics of animal neural networks and performs distributed parallel information processing. Relying on the complexity of the system, the neural network achieves the purpose of processing information by adjusting the interconnection relationship between a large number of internal computing nodes. [0003] Neural networks have been widely and successfully applied in many fields such as speech recognition, text recognition, and image and video recognition. Traditional neural networks use double-precision or single-precision floating-point multiplication / addition calculations as the basic calculation unit. How...

AI Technical Summary

Problems solved by technology

Traditional neural networks use double-precision or single-precision floating-point multiplication / addition calculations as the basic calculation unit. However, the algorithm architecture of traditional neural networks will result in large computational requirements, high memory (or video memory) occupation, and high bandwidth requirements. The problem, not only has higher requirements for hardware, but also increases the cost

Images