US 11,841,921 B2
Model training method and apparatus, and prediction method and apparatus
Xibin Song, Beijing (CN); Dingfu Zhou, Beijing (CN); Jin Fang, Beijing (CN); and Liangjun Zhang, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed by BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed on Dec. 4, 2020, as Appl. No. 17/112,247.
Claims priority of application No. 202010593221.X (CN), filed on Jun. 26, 2020.
Prior Publication US 2021/0406599 A1, Dec. 30, 2021
Int. Cl. G06T 7/50 (2017.01); G06N 20/00 (2019.01); G06F 18/214 (2023.01); G06V 10/42 (2022.01); G06T 3/40 (2006.01)
CPC G06F 18/214 (2023.01) [G06N 20/00 (2019.01); G06T 3/40 (2013.01); G06T 7/50 (2017.01); G06V 10/42 (2022.01); G06T 2207/10028 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A model training method, comprising:
inputting a first sample image of sample images into a depth information prediction model, and acquiring depth information of the first sample image;
acquiring inter-image posture information based on a second sample image of the sample images and the first sample image;
acquiring a projection image of the first sample image in a view of the second sample image, at least according to the inter-image posture information and the depth information; and
acquiring a loss function by determining a function for calculating a similarity between the second sample image and the projection image, and training the depth information prediction model using the loss function;
wherein, the depth information of the first sample image is acquired by:
concatenating an image feature and a convolutional feature of a same feature size to obtain concatenated features of multiple sizes;
determining intermediate depth information based on the concatenated features of multiple sizes; and
determining the depth information of the first sample image based on the intermediate depth information.