Hardware accelerator for supporting 2d convolution and 3D convolution operations of density tensor and sparse tensor

A versatile CNN computation accelerator supports 2D/3D convolutions and dense/sparse tensors, adapting to different applications without hardware changes, enhancing efficiency and flexibility in small embedded systems.

WO2026141739A1PCT designated stage Publication Date: 2026-07-02KOREA ELECTRONICS TECH INST

Patent Information

Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
KOREA ELECTRONICS TECH INST
Filing Date
2024-12-27
Publication Date
2026-07-02

Smart Images

  • Figure KR2024021273_02072026_PF_FP_ABST
    Figure KR2024021273_02072026_PF_FP_ABST
Patent Text Reader

Abstract

A hardware accelerator for supporting 2D convolution and 3D convolution operations of a density tensor and a sparse tensor is provided. A CNN operation accelerator according to an embodiment of the present invention is versatile hardware that can support both 2D / 3D convolution and density / sparse tensor operations without hardware changes, and thus can be used for various purposes in small embedded systems. Therefore, a need for different network models according to applications can be prevented.
Need to check novelty before this filing date? Find Prior Art