公开(公告)号：US11475276B1

公开(公告)日：2022-10-18

申请号：US15804900

申请日：2017-11-06

Applicant: Apple Inc.

Inventor： Ashish Shrivastava ,  Tomas J. Pfister ,  Cuneyt O. Tuzel ,  Russell Y. Webb ,  Joshua Matthew Susskind

IPC: G06N3/04 ,  G06N3/08 ,  G06K9/62

Abstract: A generative network may be learned in an adversarial setting with a goal of modifying synthetic data such that a discriminative network may not be able to reliably tell the difference between refined synthetic data and real data. The generative network and discriminative network may work together to learn how to produce more realistic synthetic data with reduced computational cost. The generative network may iteratively learn a function that synthetic data with a goal of generating refined synthetic data that is more difficult for the discriminative network to differentiate from real data, while the discriminative network may be configured to iteratively learn a function that classifies data as either synthetic or real. Over multiple iterations, the generative network may learn to refine the synthetic data to produce refined synthetic data on which other machine learning models may be trained.

公开(公告)号：US20240153201A1

公开(公告)日：2024-05-09

申请号：US18451242

申请日：2023-08-17

Applicant: Apple Inc.

Inventor： Anurag Ranjan ,  Kwang M. Yi ,  Cuneyt O. Tuzel

IPC: G06T15/50 ,  G06T15/08 ,  G06T17/00

CPC classification number: G06T15/50 ,  G06T15/08 ,  G06T17/00

Abstract: An electronic device may include a light based image generation system configured to generate images of a 3-dimensional object in a scene. The light based image generation system can include a feature extractor, a triplane decoder, and a volume renderer. The feature extractor can receive lighting information about the scene and a perspective of the object in the scene and generate corresponding triplane features. The triplane decoder can decode diffuse and specular reflection parameters based on the triplane features. The volume renderer can render a set of images based on the diffuse and specular reflection parameters. A super resolution image can be generated from the set of images and compared with ground truth images to fine tune weights, biases, and other machine learning parameters associated with the light based image generation system. The light based image generation system can be conditioned to generate photorealistic images of human faces.

公开(公告)号：US20230081346A1

公开(公告)日：2023-03-16

申请号：US18046871

申请日：2022-10-14

Applicant: Apple Inc.

Inventor： Ashish Shrivastava ,  Tomas J. Pfister ,  Cuneyt O. Tuzel ,  Russell Y. Webb ,  Joshua Matthew Susskind

IPC: G06N3/04 ,  G06N3/08

Abstract: A generative network may be learned in an adversarial setting with a goal of modifying synthetic data such that a discriminative network may not be able to reliably tell the difference between refined synthetic data and real data. The generative network and discriminative network may work together to learn how to produce more realistic synthetic data with reduced computational cost. The generative network may iteratively learn a function that synthetic data with a goal of generating refined synthetic data that is more difficult for the discriminative network to differentiate from real data, while the discriminative network may be configured to iteratively learn a function that classifies data as either synthetic or real. Over multiple iterations, the generative network may learn to refine the synthetic data to produce refined synthetic data on which other machine learning models may be trained.

公开(公告)号：US20210374570A1

公开(公告)日：2021-12-02

申请号：US17326098

申请日：2021-05-20

Applicant: Apple Inc.

Inventor： Joseph Y. Cheng ,  Erdrin Azemi ,  Hanlin Goh ,  Kaan E. Dogrusoz ,  Cuneyt O. Tuzel

IPC: G06N5/04 ,  G06N20/00

Abstract: The present application relates to apparatus, systems, and methods to perform subject-aware self-supervised learning of a machine-learning model for classification of data, such as classification of biosignals.

公开(公告)号：US10970518B1

公开(公告)日：2021-04-06

申请号：US16188879

申请日：2018-11-13

Applicant: Apple Inc.

Inventor： Yin Zhou ,  Cuneyt O. Tuzel ,  Jerremy Holland

IPC: G06K9/00 ,  G06T15/08 ,  G06T7/521 ,  G06K9/62 ,  G06T9/00

Abstract: A voxel feature learning network receives a raw point cloud and converts the point cloud into a sparse 4D tensor comprising three-dimensional coordinates (e.g. X, Y, and Z) for each voxel of a plurality of voxels and a fourth voxel feature dimension for each non-empty voxel. In some embodiments, convolutional mid layers further transform the 4D tensor into a high-dimensional volumetric representation of the point cloud. In some embodiments, a region proposal network identifies 3D bounding boxes of objects in the point cloud based on the high-dimensional volumetric representation. In some embodiments, the feature learning network and the region proposal network are trained end-to-end using training data comprising known ground truth bounding boxes, without requiring human intervention.