Title:
CONVOLUTIONAL NEURAL NETWORK ACCELERATION DEVICE, CONVOLUTIONAL NEURAL NETWORK ACCELERATION METHOD, AND CONVOLUTIONAL NEURAL NETWORK ACCELERATION PROGRAM
Document Type and Number:
WIPO Patent Application WO/2024/053120
Kind Code:
A1
Abstract:
This convolutional neural network acceleration device includes a model structure analyzing unit and a model transforming unit. The model structure analyzing unit analyzes an internal structure of an input CNN model, and outputs transformation target information specifying a layer in which a transformation contributing to acceleration can be performed. The model transforming unit uses input transformation specifying information and the transformation target information output from the model structure analyzing unit to transform the input CNN model, and outputs a transformed CNN model.
More Like This:
Inventors:
TAKAGI MOTOHIRO (JP)
KONDO SHIGEKUNI (JP)
SESHIMO HITOSHI (JP)
AONO YUSHI (JP)
KONDO SHIGEKUNI (JP)
SESHIMO HITOSHI (JP)
AONO YUSHI (JP)
Application Number:
PCT/JP2022/043378
Publication Date:
March 14, 2024
Filing Date:
November 24, 2022
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G06N3/082
Foreign References:
| JP2021103441A | 2021-07-15 |
Other References:
HIROTAKA NAKAHARA , HARUYOSHI YONEKAWA , TOMOYA FUJII , MASAYUKI SHIMODA , SHINPEI SATO: "About the deep learning development environment GUINNESS for FPGA", IEICE TECHNICAL REPORT, IEICE, JP, vol. 117, no. 221, 18 September 2017 (2017-09-18), JP, pages 51 - 56, XP009553107, ISSN: 0913-5685
Attorney, Agent or Firm:
SUZUYE & SUZUYE (JP)
Download PDF: