Method and system for generating training data
AssigneeVicarious FPC, Inc.
InventorsEyrun Eyjolfsdottir, Michael Stark
Life of this patent
Estimating the term…
Abstract
A method for generating training data can include: determining a set of images; determining a set of masks based on the images; determining a first mesh based on the set of masks; optionally determining a refined mesh by recomputing the first mesh; optionally determining one or more faces of the refined mesh; optionally adding one or more keypoints to the refined mesh; optionally determining a material property set for the object; optionally generating a full object mesh; determining one or more scenes; optionally determining training data based on the one or more scenes; optionally training one or more object detectors using the training data; and detecting one or more objects using the trained object detector.
Claims22
A method, comprising:
- generating an object mesh comprising a set of object components, wherein the object mesh is associated with a plurality of object keypoints, each labeled with an object component identifier for an object component of the set that the object keypoint is proximal to;
- rendering a synthetic image using the object mesh;
- associating the synthetic image with object keypoints of the plurality that are unoccluded in the synthetic image;
- and
- training an object detector to identify the unoccluded object keypoints in the synthetic image.
A method, comprising:
- capturing an image of a physical scene using an imaging system proximal the physical scene;
- and
- detecting an inference keypoint, associated with an object component identifier for an object component of an object in the physical scene using an object detector trained on training data generated by:
- determining an object mesh associated with a plurality of object keypoints that are each associated with a position relative to an object mesh of an object geometry and a respective object component identifier;
- rendering a synthetic image using the object mesh;
- and
- training the object detector to detect the object keypoints of the object mesh within the synthetic image.
Dependencies are inferred from the claim text.