Title:
DEEP RENDERING MODEL TRAINING METHOD AND APPARATUS, AND TARGET RENDERING METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2023/226479
Kind Code:
A1
Abstract:
The present disclosure relates to the field of artificial intelligence. Provided are a deep rendering model training method and apparatus, and a target rendering method and apparatus. The deep rendering model training method comprises: by means of preset control parameters, rendering a target to be rendered, so as to generate a sample image frame; inputting the control parameters into a machine learning model; inputting into the machine learning model a tensor to be processed, so as to render said tensor according to the control parameters by using the machine learning model, thereby generating a rendered image frame; according to the rendered image frame and the sample image frame, determining a loss function value; and training the machine learning model and said tensor by using the loss function value so as to generate a deep rendering model and a target tensor. The target rendering method comprises: acquiring control parameters associated with a rendering target; and inputting the control parameters into the deep rendering model so as to render the target tensor according to the control parameters by using the deep rendering model to generate an image frame.
More Like This:
Inventors:
SHEN TONG (CN)
ZHANG WEI (CN)
MEI TAO (CN)
ZHANG WEI (CN)
MEI TAO (CN)
Application Number:
PCT/CN2023/075887
Publication Date:
November 30, 2023
Filing Date:
February 14, 2023
Export Citation:
Assignee:
BEIJING JINGDONG SHANGKE INFORMATION TECHNOLOGY CO LTD (CN)
International Classes:
G06T15/20
Foreign References:
| CN115049780A | 2022-09-13 | |||
| CN113538659A | 2021-10-22 | |||
| CN112749758A | 2021-05-04 | |||
| US20170154425A1 | 2017-06-01 | |||
| US20210097730A1 | 2021-04-01 | |||
| CN113810597A | 2021-12-17 |
Attorney, Agent or Firm:
CCPIT PATENT AND TRADEMARK LAW OFFICE (CN)
Download PDF: