Quantization apparatus and method for artificial neural network and image processing device having the same

An artificial neural network-based image processing device, comprising:

• a memory;
• and
• a processor that executes at least a part of an operation according to a neural network model stored in the memory,
• wherein the processor converts a received image to obtain an embedding matrix, quantizes the embedding matrix and a plurality of weight matrices obtained by training, respectively, performs a neural network operation on the quantized embedding matrix and the quantized weight matrices to identify a class,
• receives one of the embedding matrix or a feature map, which is a matrix obtained by matrix multiplication between the embedding matrix and the weight matrices, as an input matrix, divides the input matrix into a plurality of channels, selects a quantizer for each channel by analyzing a distribution of element values included in each of the plurality of divided channels, and quantizes element values included in a channel using the selected quantizer.

A quantization method, which is performed by a processor executing at least part of an operation according to a neural network model,

• wherein the method includes the steps of:
• receiving one of a plurality of matrices obtained for neural network operation in an artificial neural network as an input matrix, and dividing the input matrix into a plurality of channels;
• selecting a quantizer for each channel by analyzing a distribution of element values included in each of the plurality of divided channels;
• and
• quantizing element values included in the channels using the selected quantizer.