Inspur Suzhou Intelligent Technology Co., Ltd

Wenfeng Yin

Yaqian Zhao

Gang Dong

Lingyan Liang

Qichun Cao

Haiwei Liu

Hongbin Yang

An image recognition method includes: inputting a sample image data set into an original neural network model in advance; and for each convolution layer of the original neural network model, using a feature map of the sample image data set in the current layer as a reconstruction target, firstly obtaining an update weight of a convolution kernel by using a kernel set construction method, then calculating an input channel combination having a minimum reconstruction error, cutting a redundant input channel to obtain a compression result of the current convolution layer, and finally, splicing compression results of the convolution layers to generate an image recognition model. An image to be recognized is obtained and is inputted into the image recognition model, and an output result of the image recognition model is used as an image recognition result of the image.

Image recognition method and device, and computer-readable storage medium

A neural network compression method whereby forward inference is performed on target data by using a target parameter sharing network to obtain an output feature map of the last convolutional module, a channel related feature is extracted from the output feature map, the extracted channel related feature and a target constraint condition are input into a target meta-generative network, and an optimal network architecture under the target constraint condition is predicted by using the target meta-generative network to obtain a compressed neural network model.

Neural network compression method, apparatus and device, and storage medium

A hardware environment-based data quantization method includes: parsing a model file under a current deep learning framework to obtain intermediate computational graph data and weight data that are independent of a hardware environment; performing calculation on image data in an input data set through a process indicated by an intermediate computational graph to obtain feature map data; separately performing uniform quantization on the weight data and the feature map data of each layer according to a preset linear quantization method, and calculating a weight quantization factor and a feature map quantization factor (S); combining the weight quantization factor and the feature map quantization factor to obtain a quantization parameter that makes hardware use shift instead of division; and finally, writing the quantization parameter and the quantized weight data to a bin file according to a hardware requirement so as to generate quantized file data (S).

Growing community of inventors

Jiangsu, China

Wenfeng Yin