A partial binary convolution method suitable for embedded devices

Abstract

The invention discloses a partial binary convolution method suitable for embedded equipment, belonging to a model compression method of depth learning. The method comprises the following steps: 1. foreach convolution layer of a given CNN, measuring the importance of each convolution core according to the statistics of each output characteristic map; the convolution kernels of each layer being divided into two groups. 2, rearrangeing that two groups of convolution core on the storage space so that the storage positions of the same group of convolution cores are adjacent to each other, recording the rearrange order to generate a new convolution layer; 3, changing the channel order of the convolution nucleus of the next convolution layer according to the rearrangement order of the step 2; Step 4, according to the CNN processed in the above steps, fine-tuning training being carried out, binary quantization being carried out on the convolution layer divided into non-important layers, and the accuracy of the whole network being gradually recovered through the iterative operation of quantization and training. The invention ensures the accuracy of the given CNN on the large data set, andreduces the calculation and storage overhead of the convolution neural network on the embedded equipment.

Description

technical field [0001] The invention belongs to the field of model compression of deep learning, and in particular relates to a partial binary convolution method suitable for embedded devices. Background technique [0002] With the rapid development of deep convolutional neural networks (CNNs), CNNs have become state-of-the-art in many task domains, such as image classification, object recognition, and semantic segmentation. Traditionally, a data center equipped with a high-end image processor (GPU) is the best choice for deploying CNN applications, but this cloud-centric application framework usually causes some problems, such as user privacy issues, long response time, These apps cannot even be used without the internet. Therefore, people started to deploy CNN directly on embedded devices. [0003] CNN models often require a lot of computing resources and storage resources. For example, AlexNet with 5 convolutional layers requires 722M floating point number operations (F...

AI Technical Summary

Problems solved by technology

This is mainly because: Binary quantization is performed uniformly for all convolution kernels without considering the different importance of different convolution kernels. If important convolution kernels are binary quantized, it may lead to a decrease in accuracy.

Images