US 12,315,180 B2
Depth estimation method and apparatus using learning model
Soon-heung Jung, Daejeon (KR); Jeongil Seo, Daejeon (KR); David Crandall, Bloomington, IN (US); and Md Alimoor Reza, Bloomington, IN (US)
Assigned to ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE, Daejeon (KR); and THE TRUSTEES OF INDIANA UNIVERSITY, Indianapolis, IN (US)
Filed by Electronics and Telecommunications Research Institute, Daejeon (KR); and The Trustees of Indiana University, Indianapolis, IN (US)
Filed on Nov. 26, 2021, as Appl. No. 17/535,921.
Prior Publication US 2023/0169671 A1, Jun. 1, 2023
Int. Cl. G06K 9/00 (2022.01); G06T 7/593 (2017.01); G06V 10/77 (2022.01)
CPC G06T 7/593 (2017.01) [G06V 10/7715 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 13 Claims
OG exemplary drawing
 
1. A depth information estimation method comprising:
identifying a plurality of image data;
generating feature maps of the plurality of image data respectively;
generating a cost volume using the feature maps;
generating normalized cost volumes having different resolutions from each other by normalizing the cost volume;
estimating disparity information from the normalized cost volumes; and
generating depth information using the estimated disparity information,
wherein the generating of the normalized cost volumes in different sizes comprises:
extracting a feature map of the cost volume from the cost volume;
downsampling the feature map of the cost volume; and
generating the normalized cost volumes in different sizes by upsampling the downsampled feature map of the cost volume, and
wherein the generating of the cost volume comprises generating the cost volume by processing the feature maps using a spatial pyramid pooling module.