Title:
METHOD FOR TRAINING TARGET RECOGNITION MODEL, TARGET RECOGNITION METHOD, APPARATUS, AND ROBOT
Document Type and Number:
WIPO Patent Application WO/2019/148453
Kind Code:
A1
Abstract:
The present application discloses a method for training a target recognition model, a target recognition method, an apparatus, and a robot. The method for training a target recognition model comprises: obtaining a sample image of a target to be recognized; inputting the sample image into a recognition model, and outputting post-recognition segmentation data; comparing the segmentation data with standard data of a standard image to obtain a recognition error; feeding back the recognition error to the recognition model, and correcting the recognition model, wherein when a standard image is obtained, the edges of the target are marked with a fluorescent material, and the standard data of the target to be recognized is obtained according to the resultant color of the fluorescent material in the standard image. The present application improves speed and efficiency of model training.
Inventors:
ZHANG HAO (CN)
WU QIFAN (CN)
WU QIFAN (CN)
Application Number:
PCT/CN2018/075134
Publication Date:
August 08, 2019
Filing Date:
February 02, 2018
Export Citation:
Assignee:
SHENZHEN DORABOT INC (CN)
International Classes:
G06V10/44; G06V10/56
Foreign References:
| CN106874914A | 2017-06-20 | |||
| CN101573959A | 2009-11-04 | |||
| CN1738426A | 2006-02-22 | |||
| CN101354359A | 2009-01-28 | |||
| CN105678332A | 2016-06-15 | |||
| CN106023220A | 2016-10-12 |
Attorney, Agent or Firm:
CHINA WISPRO INTELLECTUAL PROPERTY LLP. (CN)
Download PDF:
Previous Patent: SPEAKER ASSEMBLY AND ELECTRONIC DEVICE
Next Patent: CHILD-PROOF RE-SEALABLE CONTAINER
Next Patent: CHILD-PROOF RE-SEALABLE CONTAINER