US 11,854,204 B2
Information processing device, information processing method, and computer program product
Shunsuke Sakurai, Kawasaki (JP); Ryo Nakashima, Kawasaki (JP); and Akihito Seki, Yokohama (JP)
Assigned to KABUSHIKI KAISHA TOSHIBA, Tokyo (JP)
Filed by KABUSHIKI KAISHA TOSHIBA, Tokyo (JP)
Filed on Feb. 26, 2021, as Appl. No. 17/186,230.
Claims priority of application No. 2020-116609 (JP), filed on Jul. 6, 2020.
Prior Publication US 2022/0004816 A1, Jan. 6, 2022
Int. Cl. G06T 7/00 (2017.01); G06T 7/11 (2017.01); G06N 3/08 (2023.01); G06T 7/13 (2017.01); G06F 18/214 (2023.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)
CPC G06T 7/11 (2017.01) [G06F 18/2155 (2023.01); G06N 3/08 (2013.01); G06T 7/13 (2017.01); G06V 10/7753 (2022.01); G06V 10/82 (2022.01)] 15 Claims
OG exemplary drawing
 
1. An information processing device comprising:
one or more processors configured to:
receive input data including first input data belonging to a first domain and second input data belonging to a second domain different from the first domain;
input data based on the input data to a first model, and acquire first output data indicating an execution result of a first task with the first model;
input data based on the input data to a second model, and acquire second output data indicating an execution result of a second task with the second model;
convert the first output data into first conversion data expressed in a form of an execution result of the second task, the execution result of the second task including a plurality of elements;
generate supervised data of the second model for the first input data, based on the first conversion data and the second output data, and
learn the second model using the first input data and the supervised data, wherein
the first task is one of a plurality of tasks including depth estimation, semantic segmentation, instance segmentation, panoptic segmentation, object detection, autoencoder, super resolution, edge detection, generation of a saliency map, and normal line estimation;
the second task is another one of the plurality of tasks;
the first model outputs the execution result of the first task;
the second model outputs the execution result of the second task; and
the supervised data includes
one or more elements having a degree of reliability larger than those of another element in both the first conversion data and the second output data, or
one or more elements being classified into a same class in both the first conversion data and the second output data.