Title:
HETEROGENEOUS COMPUTING DEVICE-ORIENTED DEEP LEARNING IMAGE CLASSIFICATION
Document Type and Number:
WIPO Patent Application WO/2024/060810
Kind Code:
A1
Abstract:
The present disclosure relates to heterogeneous computing device-oriented deep learning image classification. According to embodiments of the present disclosure, a deep learning model can be modeled as an initial directed acyclic graph, nodes represent operators of the deep learning model, and directed edges represent data transmission among the operators. Then, the directed edges in the initial directed acyclic graph are replaced with new nodes, and two directed edges are added to maintain the topology, so as to generate a new directed acyclic graph. In this way, constraint conditions can be constructed for memory occupation of the operators, the memory limit of a device, the execution order of the operators on the basis of the initial directed acyclic graph and the new directed acyclic graph; and on the basis of the processing time of the operators on each device and the transmission time of data among the operators between devices and with the minimization of the inference completion time of the deep learning model as an objective, the operators of the deep learning model are allocated to a plurality of devices for execution.
More Like This:
| WO/2019/006228 | METHOD FOR TREATING A SURFACE |
| WO/2020/081470 | DEEP LEARNING INFERENCE SYSTEMS AND METHODS FOR IMAGING SYSTEMS |
| WO/2024/074921 | DISTINGUISHING A DISEASE STATE FROM A NON-DISEASE STATE IN AN IMAGE |
Inventors:
ZHANG BEIBEI (CN)
GAO FENG (CN)
SUN MINGGE (CN)
ZHENG CHU (CN)
GAO FENG (CN)
SUN MINGGE (CN)
ZHENG CHU (CN)
Application Number:
PCT/CN2023/107827
Publication Date:
March 28, 2024
Filing Date:
July 18, 2023
Export Citation:
Assignee:
ZHEJIANG LAB (CN)
International Classes:
G06V10/764
Foreign References:
| CN115249315A | 2022-10-28 | |||
| CN112297014A | 2021-02-02 | |||
| CN109561148A | 2019-04-02 | |||
| CN112381211A | 2021-02-19 | |||
| US20030085932A1 | 2003-05-08 |
Other References:
NA LI, GAO BO; XIE ZONGFU: "Research on Scheduling Method of Layered Heterogeneous Signal Processing Platform", DIANZI KEJI =IT AGE, XI'AN DIANZI KEJI DAXUE, CN, vol. 35, no. 2, 15 February 2022 (2022-02-15), CN , pages 7 - 13, XP093149613, ISSN: 1007-7820, DOI: 10.16180/j.cnki.issn1007-7820.2022.02.002
Attorney, Agent or Firm:
BEIJING BESTIPR INTELLECTUAL PROPERTY LAW CORPORATION (CN)
Download PDF: