公开(公告)号：US20220353458A1

公开(公告)日：2022-11-03

申请号：US17746929

申请日：2022-05-17

Applicant: Apple Inc.

Inventor： Peter ZATLOUKAL

IPC: H04N7/01 ,  H04N5/917 ,  H04N7/18 ,  G06K9/62 ,  G06V20/40

Abstract: In one embodiment, a computing device receives, from one or more cameras, a video stream comprising multiple frames, where the video stream is received at a first quality. The computing device analyzes, using a machine-learning model, images in the frames, where the machine-learning model has been trained to detect one or more objects-of-interest in the images. The computing device identifies a sequence-of-interest including consecutive frames of the video stream, where at least one object-of-interest was detected in at least one of the consecutive frames. The computing device generates a video package including the sequence-of-interest.

公开(公告)号：US20250037018A1

公开(公告)日：2025-01-30

申请号：US18658919

申请日：2024-05-08

Applicant: Apple Inc.

Inventor： Minsik CHO ,  Keivan ALIZADEH VAHID ,  Qichen FU ,  Saurabh ADYA ,  Carlo Eduardo Cabanero DEL MUNDO ,  Mohammad RASTEGARI ,  Devang K. NAIK ,  Peter ZATLOUKAL

IPC: G06N20/00

Abstract: The subject technology provides memory-efficient differentiable weight clustering for large language model compression. An apparatus determines a tensor including an attention map between learned weights of a trained machine learning model and corresponding centroids. The apparatus also determines a compressed attention table and a plurality of index lists during compression of the trained machine learning model based on an uniquification of the attention map and sharding of an associated index list. The apparatus determines whether the tensor exists at a destination device during compression of the trained machine learning model using a marshaling layer. The apparatus refrains from copying the tensor to the destination device when the tensor exists at the destination device, or copies the tensor to the destination device when the tensor does not exist at the destination device. The apparatus deploys a compressed machine learning model based on the compression of the trained machine learning model.

公开(公告)号：US20240119699A1

公开(公告)日：2024-04-11

申请号：US18485266

申请日：2023-10-11

Applicant: Apple Inc.

Inventor： Hessam BAGHERINEZHAD ,  Carlo Eduardo Cabanero DEL MUNDO ,  Anish Jnyaneshwar PRABHU ,  Peter ZATLOUKAL ,  Lawrence Frederick ARNSTEIN

IPC: G06V10/44 ,  G06F18/214 ,  G06N20/00 ,  G06T7/00 ,  G06T7/20 ,  G06V10/764 ,  G06V10/82 ,  G06V20/52 ,  G06V40/20

CPC classification number: G06V10/454 ,  G06F18/214 ,  G06N20/00 ,  G06T7/0002 ,  G06T7/20 ,  G06V10/764 ,  G06V10/82 ,  G06V20/52 ,  G06V40/20 ,  G06T2207/20081 ,  G06V20/44

Abstract: In one embodiment, a method includes receiving an input video comprising a plurality of image frames including an object of interest. Based on the plurality of image frames, a motion associated with the object of interest is determined, and the plurality of image frames are classified using a machine-learning model to identify one of the plurality of image frames that indicates a moment of perception of the determined motion.