Title:
MULTI-LAYER FEATURE FUSION-BASED ENDPOINT DETECTION METHOD, APPARATUS, DEVICE, AND MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/135281
Kind Code:
A1
Abstract:
A multi-layer feature fusion-based endpoint detection method, apparatus, device, and a medium. The method comprises: separately inputting a sample dataset having a label into an encoder of an initial endpoint detection network, and inputting outputs of the encoder into a decoder; fusing features extracted by the encoder with features that are extracted by the decoder and that have the same definition, in order to obtain fused features of each piece of sample data; using the fused features of each piece of the sample data to determine an endpoint predicted value of each piece of the sample data, and using a loss function to calculate the difference between the endpoint predicted value of each piece of the sample data and a label value; using a reverse propagation algorithm to optimize parameters of the initial endpoint detection network until a preset precision is achieved, and obtaining a target endpoint detection network; and inputting data to be detected into the target endpoint detection network, and outputting an endpoint detection result. The method, apparatus, device, and computer-readable medium provided by the present invention improve the adaptiveness and robustness of endpoint detection.
Inventors:
LI FENG (CN)
LIU HONGGANG (CN)
LIU HONGGANG (CN)
Application Number:
PCT/CN2020/110977
Publication Date:
July 08, 2021
Filing Date:
August 25, 2020
Export Citation:
Assignee:
INSPUR BEIJING ELECTRONIC INF IND CO LTD (CN)
International Classes:
H03M13/25
Domestic Patent References:
| WO2008069519A1 | 2008-06-12 |
Foreign References:
| CN111181574A | 2020-05-19 | |||
| CN106971230A | 2017-07-21 | |||
| CN109919013A | 2019-06-21 | |||
| CN109801646A | 2019-05-24 | |||
| CN109685127A | 2019-04-26 |
Attorney, Agent or Firm:
UNITALEN ATTORNEYS AT LAW (CN)
Download PDF: