Quantitative perception training method and device, equipment, medium and convolutional neural network

The invention discloses a quantitative perception training method and device, equipment, a medium and a convolutional neural network. The method is executed by the electronic equipment and comprises the steps that sample training is carried out on a first convolutional neural network, and at least one convolutional layer in the first convolutional neural network comprises a mergeable branch structure and at least one first direct connection branch; at least one first direct connection branch in at least one convolutional layer of the trained first convolutional neural network is reserved, and combinable branch structures except the first direct connection branch in the convolutional layers are combined to obtain a second convolutional neural network with a first direct connection branch structure, and performing quantitative perception training based on the second convolutional neural network to improve the quantization precision of the convolutional neural network model. 本申请公开了一种量化感知训练方法、装置、设备、介质及卷积神经网络。方法由电