ACQUISITION METHOD FOR TRAINING SAMPLE SET, MODEL TRAINING METHOD AND RELATED APPARATUS

WIPO Patent Application WO/2023/015880

A1

Disclosed in the present application are an acquisition method for a training sample set, a model training method and a related apparatus. The acquisition method comprises: acquiring several image groups, and determining an image difference degree between a first dense depth map and a first sparse depth map in each image group; and selecting an image group, the image difference degree of which meets a preset condition, from among the several image groups, so as to obtain a training sample set. In the present application, dense depth maps are acquired by means of a binocular camera, image difference degrees between the dense depth maps and sparse depth maps, which are acquired by means of a distance measurement apparatus, are determined, and an image group, the image difference degree of which meets a preset condition, is taken as a training sample. In this way, by using the dense depth maps acquired by means of the binocular camera, the reliability of the dense depth maps can be ensured; and the dense depth maps are screened by means of the sparse depth maps acquired by means of the distance measurement apparatus, and a dense depth map having a large deviation from an actual scenario is eliminated to improve the accuracy of a dense depth map serving as a ground truth, thereby improving the model performance of a depth information completing model which is obtained by performing training on the basis of a training sample set.

WO/2023/033697	A COMPUTER-IMPLEMENTED OR HARDWARE-IMPLEMENTED METHOD FOR PROCESSING DATA, A COMPUTER PROGRAM PRODUCT, A DATA PROCESSING SYSTEM AND A FIRST CONTROL UNIT THEREFOR
WO/2022/257329	BRAIN MACHINE INTERFACE DECODING METHOD BASED ON SPIKING NEURAL NETWORK
WO/2023/063874	METHOD AND SYSTEM FOR IMAGE PROCESSING BASED ON CONVOLUTIONAL NEURAL NETWORK

LIU LIU (CN)
XU YUHUA (CN)
YAN MIN (CN)
YU YU SHAN (CN)
YANG XIAOLI (CN)
ZHAO XIN (CN)

PCT/CN2022/080515

February 16, 2023

March 13, 2022

Click for automatic bibliography generation  

ORADAR TECH COMPANY LIMITED (CN)

G06N3/04

WO2021013334A1

2021-01-28

CN113780349A	2021-12-10
CN112541482A	2021-03-23
CN112330729A	2021-02-05
CN110148086A	2019-08-20

SHENZHEN HUNS INTELLECTUAL PROPERTY (CN)

View/Download PDF      PDF Help

Previous Patent: ELASTIC SUPPORT MEMBER, ELECTROLYTIC CELL, MANUFACTURING APPARATUS AND MANUFACTURING METHOD

Next Patent: PATTERN INFORMATION ACQUISITION METHOD AND APPARATUS, ELECTRONIC DEVICE, AND READABLE STORAGE MEDIUM