Convolutional neural network (CNN) processing method and apparatus

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published April 7th, 2026

Detected April 7th, 2026

Summary

The USPTO granted Samsung Electronics Co., Ltd. patent US12596913B2 for a convolutional neural network (CNN) processing apparatus and method. The patent covers techniques for determining loading space units in input feature maps, loading target input elements from memory into buffers, and performing convolution operations with kernel elements. The patent contains 13 claims and names Jinwoo Son, Changyong Son, Jaejoon Han, and Chang Kyu Choi as inventors.

View original document View source feed page

What changed

The USPTO issued a patent grant (B2) to Samsung Electronics Co., Ltd. for US12596913B2, a convolutional neural network processing apparatus and method. The invention relates to determining loading space units based on input feature map dimensions and kernel dimensions, loading target input elements from memory into allocated buffers, and performing convolution operations. The patent was filed on October 28, 2022, with application number 17975837.

For Samsung Electronics, this patent grant establishes enforceable intellectual property rights in CNN processing technology that can be used for licensing negotiations, asserting against potential infringers, or blocking competitors from using similar methods. Third parties developing CNN hardware accelerators, AI processors, or neural network processing units should review this patent to assess freedom-to-operate implications and consider potential licensing needs.

What to do next

Monitor for updates

Source document (simplified)

← USPTO Patent Grants

Convolutional neural network (CNN) processing method and apparatus

Grant US12596913B2 Kind: B2 Apr 07, 2026

Assignee

Samsung Electronics Co., Ltd.

Inventors

Jinwoo Son, Changyong Son, Jaejoon Han, Chang Kyu Choi

Abstract

Disclosed is a convolutional neural network (CNN) processing apparatus and method, the apparatus configured to determine a loading space unit for at least one loading space in an input based on a height or a width for an input feature map of the input and an extent of a dimension of a kernel feature map, load target input elements corresponding to a target loading space, among the at least one loading space, from a memory and store the target input elements in an allocated input buffer having a size corresponding to the loading space unit, and perform a convolution operation between the target input elements stored in the input buffer and at least one kernel element of a kernel.