Convolutional neural network (CNN) processing method and apparatus

A convolutional neural network (CNN) processing apparatus comprising:

• one or more processors configured to implement a convolutional layer of a neural network, through the one or more processors being configured to:
• acquire at least one convolution operation result between at least one kernel element and at least one input element of an input of a convolutional layer, wherein the at least one input element is loaded from a memory storing input elements interleaved by channel according to a horizontal or vertical rasterizing scheme to reduce overlapping data loads during convolution operations;
• and
• generate an output of the convolutional layer based on the at least one operation result by mapping the at least one operation result to an output feature map to which padding is applied based on a size of a pad predetermined to correspond to an input of a subsequent convolutional layer of the convolutional layer in the neural network,
• wherein the generating of the output in consideration of the size of the pad enables skipping of operations on the applied padding when performing convolution operations of the subsequent convolutional layer based on kernel information indicating skip targets pre-classified from kernel elements, thereby reducing a number of the convolution operations and computational resources required for processing the subsequent convolutional layer.