Convolutional neural network processing method and device, equipment and storage medium

Abstract

The invention discloses a convolutional neural network processing method and device, equipment and a storage medium. The method comprises: obtaining an original weight matrix and an original input neuron matrix of the convolutional neural network, sequentially carrying out Winograd transformation and quantization processing on the original weight matrix to obtain a target weight matrix, sequentially carrying out quantization processing and Winograd transformation on the original input neuron matrix to obtain a target input neuron matrix, and obtaining an output neuron matrix of the convolutional neural network according to the target weight matrix and the target input neuron matrix. According to the embodiment of the invention, the time complexity of the convolutional neural network is reduced through Winograd transformation, and the data bit width of the convolutional neural network is reduced through quantization processing, so that the calculation speed of the convolutional neural network is improved.

Description

technical field [0001] Embodiments of the present invention relate to deep learning technology, and in particular to a convolutional neural network processing method, device, device, and storage medium. Background technique [0002] Since AlexNet was proposed in 2012, the convolutional neural network has achieved great success in the field of image processing. In major image competitions, the effect of the convolutional neural network far exceeds the traditional algorithm, and frequently refreshes various evaluation indicators in the industry. [0003] Due to the requirements of information security and low latency, the calculation of the neural network needs to be migrated from the cloud to the mobile terminal. However, with the improvement of the effect of the convolutional neural network, the model of the convolutional neural network becomes more and more complex, and the amount of calculation increases sharply. In the cloud, the convolutional neural network can be accele...

AI Technical Summary

Problems solved by technology

[0003] Due to the requirements of information security and low latency, the calculation of the neural network needs to be migrated from the cloud to the mobile terminal. However, with the improvement of the effect of the convolutional neural network, the model of the convolutional neural network becomes more and more complex, and the amount of calculation increases sharply.

In the cloud, the convolutional neural network can be accelerated by relying on GPU (Graphics Processing Unit) parallel computing, while in the mobile terminal, due to the relative lack of computing resources, the calculation speed of the convolutional neural network cannot be improved, and thus Unable to realize real-time operation on mobile terminals

Images