Title:
TARGET DETECTION METHOD AND APPARATUS BASED ON FEDERATED LEARNING, AND DEVICE AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/189906
Kind Code:
A1
Abstract:
A target detection method based on federated learning, the method comprising: performing pruning processing on an initial target detection model by using a random weight so as to obtain a lightweight target detection model (S1); training the lightweight target detection model on the basis of a local data set so as to obtain a plurality of corresponding model gradient parameters, and sending the model gradient parameters to a server (S2); receiving a global gradient parameter obtained by fusing of the model gradient parameters by the server according to a federated averaging algorithm (S3); updating the initial target detection model by using the global gradient parameter (S4), and returning to the training step until a pre-set termination condition is met (S5), so as to obtain a target detection model; and performing, by using the target detection model, target detection on an image to be subjected to detection (S6). In addition, the present application further relates to blockchain technology, a local data set can be stored in a blockchain node, and by means of the method, the detection speed and detection accuracy of target detection can be improved.
Inventors:
LI ZEYUAN (CN)
WANG JIANZONG (CN)
WANG JIANZONG (CN)
Application Number:
PCT/CN2020/131986
Publication Date:
September 30, 2021
Filing Date:
November 27, 2020
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G06K9/62; G06N3/04; G06N3/08
Foreign References:
CN110309847A | 2019-10-08 | |||
CN111598238A | 2020-08-28 | |||
CN111461291A | 2020-07-28 | |||
CN111310932A | 2020-06-19 | |||
CN111126627A | 2020-05-08 | |||
CN111382706A | 2020-07-07 | |||
US20200258006A1 | 2020-08-13 |
Attorney, Agent or Firm:
SHENZHEN WORLD INTELLECTUAL PROPERTY AGENCY (GENERAL PARTNERSHIP) (CN)
Download PDF: