Convolutional neural network pruning and reasoning method and device and computer readable medium

The embodiment of the invention provides a convolutional neural network pruning and reasoning scheme, and the scheme can determine the significance metric value of each group of convolution kernels in the ith convolutional layer in a convolutional neural network model, then carries out the pruning processing of at least one group of convolution kernels with the minimum significance metric value in the ith convolutional layer, and on the basis of the i < th > convolutional layer after pruning processing, updates the feature map and the convolution kernel of the (i + 1) < th > convolutional layer. The significance metric value represents the degree of influence of the convolution kernel on obtaining of a correct result by the convolutional neural network model; after the convolution kernel with the minimum significance metric value and the convolution kernel associated with the next layer are cut off, the calculation amount of the convolution layer during reasoning can be reduced under the condition that the in