Semantic guidance for 3D reconstruction
AssigneeHuawei Technologies Co., Ltd.
InventorsSato Hiroyuki, Uragaki Keishiro
Life of this patent
Estimating the term…
Abstract
A device for reconstructing a 3D model is disclosed. The device includes: a camera; and a processor configured to: determine a class of a subject; capture one or more images of the subject by using the camera; extract semantic information from the one or more images based on the determined class; update the 3D model by using the one or more images; estimate incomplete semantic parts of the 3D model; and present semantic guidance for a user to capture the incomplete semantic parts. In this way, clearly understandable guidance is provided for reconstructing a complete 3D model.
Claims20
A device for reconstructing a 3D model, comprising:
- a camera;
- and
- at least one processor configured to:
- determine a class of a subject;
- capture one or more images of the subject by using the camera;
- extract semantic information from the one or more images based on the determined class of the subject;
- update the 3D model by using the one or more images;
- generate a semantic 3D model, the semantic 3D model being a 3D model with the semantic information;
- deform a template model to the semantic 3D model with the semantic information to generate a deformed template model;
- estimate one or more incomplete semantic parts of the 3D model based on the semantic 3D model and the deformed template model;
- and
- present semantic guidance for a user to capture the one or more incomplete semantic parts.
A device for reconstructing a 3D model, comprising:
- a camera;
- and
- at least one processor configured to:
- determine a class of a subject;
- capture one or more images of the subject by using the camera;
- update the 3D model by using the one or more images;
- deform a template model to a semantic 3D model with semantic information to generate a deformed template model;
- calculate an error for each semantic part between the semantic 3D model and the deformed template model;
- determine one or more incomplete semantic parts based on the error;
- and
- present semantic guidance for a user to capture the one or more incomplete semantic parts.
A method for reconstructing a 3D model, comprising:
- determining a class of a subject;
- capturing one or more images of the subject by using a camera;
- extracting semantic information from the one or more images based on the determined class of the subject;
- updating the 3D model by using the one or more images;
- generating a semantic 3D model, the semantic 3D model being a 3D model with the semantic information;
- deforming a template model to the semantic 3D model with the semantic information to generate a deformed template model;
- estimating one or more incomplete semantic parts of the 3D model based on the semantic 3D model and the deformed template model;
- and
- presenting semantic guidance for a user to capture the one or more incomplete semantic parts.
A non-transitory storage media storing a program thereon, when the program is executed by at least one processor, the program makes at least one processor to perform operations comprising:
- determining a class of a subject;
- capturing one or more images of the subject by using a camera;
- extracting semantic information from the one or more images based on the determined class;
- updating a 3D model by using the one or more images;
- generating a semantic 3D model, the semantic 3D model being a 3D model with the semantic information;
- deforming a template model to the semantic 3D model with the semantic information to generate a deformed template model;
- estimating one or more incomplete semantic parts of the 3D model based on the semantic 3D model and the deformed template model;
- and
- presenting semantic guidance for a user to capture the one or more incomplete semantic parts.
Dependencies are inferred from the claim text.