US 12,444,171 B2
Systems and methods for annotating 3D data
Meng Zheng, Cambridge, MA (US); Srikrishna Karanam, Bangalore (IN); Ziyan Wu, Lexington, MA (US); Arun Innanje, Lexington, MA (US); and Terrence Chen, Lexington, MA (US)
Assigned to Shanghai United Imaging Intelligence Co., Ltd., Shanghai (CN)
Filed by Shanghai United Imaging Intelligence Co., Ltd., Shanghai (CN)
Filed on Oct. 20, 2022, as Appl. No. 17/969,876.
Prior Publication US 2024/0135684 A1, Apr. 25, 2024
Prior Publication US 2024/0233338 A9, Jul. 11, 2024
Int. Cl. G06V 10/774 (2022.01); G06T 7/00 (2017.01); G06V 10/22 (2022.01); G06V 20/70 (2022.01)
CPC G06V 10/774 (2022.01) [G06T 7/0012 (2013.01); G06V 10/235 (2022.01); G06V 20/70 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20108 (2013.01)] 15 Claims
OG exemplary drawing
 
1. An apparatus, comprising:
at least one processor configured to:
obtain a first sequence of two-dimensional (2D) images associated with a first three-dimensional (3D) image dataset;
receive a 2D annotation of an object of interest associated with a first 2D image of the first sequence of 2D images;
generate a first 3D annotation of the object of interest based at least on the received 2D annotation of the object of interest, wherein the first 3D annotation is generated by automatically annotating, based on the received 2D annotation of the object of interest and a first machine-learned (ML) data annotation model, multiple other 2D images of the first sequence of 2D images, the first ML data annotation model pre-trained for extracting respective features associated with the object of interest from the multiple other 2D images based on the received 2D annotation and automatically annotating the multiple other 2D images based on the extracted features;
obtain a second 3D image dataset; and
generate a second 3D annotation of the object of interest based at least on the first 3D annotation of the object of interest, the second 3D image dataset, and a second ML data annotation model, wherein the second ML data annotation model has been pre-trained for determining a similarity between the first 3D image dataset and the second 3D image dataset and automatically annotating a second 2D image associated with the second 3D image dataset based on the determined similarity.