A Convolutional Neural Network Accelerated Computing Method for Image Super-resolution

Abstract

The invention discloses an image super-resolution convolutional neural network acceleration calculation method, which includes the following steps: (1) obtaining a trained convolution kernel group, (2) converting the trained convolution kernel group into an easy-to-use The convolution kernel group in the form of a matrix for convolution calculation processing; (3) parse the convolution kernel group in the matrix form of the intermediate convolution layer as the original convolution kernel group; (4) construct a low-rank learning model based on the original convolution kernel group : (5) Solve the base convolution kernel group by the low-rank learning model; (6) Solve the reconstruction coefficient by the least squares model; (7) Use the convolution calculation of the original convolution kernel group by using the base convolution kernel group and the corresponding The reconstruction coefficients are equivalently replaced by convolution calculations to accelerate convolution calculations. The invention reconstructs the convolution kernel group to ensure that the convolution calculation acceleration is realized without reducing the accuracy rate, and the method only involves the convolution calculation process without changing the original accuracy rate, and can be further combined with other acceleration methods Accelerate further.

Description

technical field [0001] The invention relates to the field of convolutional neural network (CNNs) accelerated computing performance optimization in artificial intelligence technology, in particular to an image super-resolution convolutional neural network accelerated computing method. Background technique [0002] At present, convolutional neural networks have become a popular technical means in the fields of computer vision, natural language processing, and speech recognition, and have achieved great technological breakthroughs. However, in the process of image super-resolution operation, due to the computational complexity of the convolution operation itself and the increasing number of network layers, the calculation of the convolution layer requires a lot of time, which makes the use of convolutional neural networks to achieve image super-resolution. rate brings great challenges. Although the convolutional operation framework designed based on GPU clusters can largely me...

AI Technical Summary

Problems solved by technology

However, in the process of image super-resolution operation, due to the complexity of the calculation of the convolution operation itself and the deeper and deeper network layers, the calculation of the convolution layer takes a lot of time. rate poses a huge challenge

Although the convolutional computing framework designed based on the GPU cluster can largely meet the training requirements of the deep convolutional neural network, the application phase is usually based on mobile terminals or embeddable devices with limited local computing capabilities. It is difficult to meet the computing needs of large-scale networks, and at the same time, the shortest possible computing time is required in the application stage. For example, the super-resolution of pictures on the mobile phone has very high real-time requirements.

Therefore, the problem of accelerating convolutional computing has been restricting the application of convolutional neural networks.

Its shortcoming is that this method directly obtains low-rank sub-tensors and factor matrices through the tensor decomposition form of the original convolutional layer, which is a process of tensor approximation and cannot completely reconstruct the original convolution. layer, it is bound to cause a reduction in accuracy, especially if the effect reduction is more obvious for image super-resolution applications

Its shortcoming is that in this method, in the low-rank decomposition operation, the low-rank decomposition of the k*k convolution kernel in the original convolution layer is decomposed into two one-dimensional convolution kernels of k*1 and 1*k to perform convolution sequentially. In this way, since the reconstruction error of the original convolution kernel is not considered, the result after convolution is quite different from the original, and it needs to be retrained later to remove redundant connections in the network, and perform k-means clustering on the weights of the remaining connections As well as steps such as encoding the number of layers, although the acceleration is more obvious, the steps are too cumbersome, and the above method is a convolution acceleration for handwritten Chinese recognition. If the image super-resolution rate is applied, the accuracy rate will drop significantly.

The currently disclosed patents are not able to accelerate the convolution calculation for image super-resolution without reducing the PSNR index.

Images