公开(公告)号：US20250061691A1

公开(公告)日：2025-02-20

申请号：US18451014

申请日：2023-08-16

Applicant: Netskope, Inc.

Inventor： Jason B. Bryslawskyj ,  Yi Zhang ,  Ari Azarafrooz ,  Wayne Xin ,  Yihua Liao ,  Niranjan Koduri ,  Emanoel Daryoush

IPC: G06V10/774 ,  G06V10/74 ,  G06V10/762 ,  G06V10/764

Abstract: Image fingerprints (embeddings) are generated by an image fingerprinting model and indexed with an approximate nearest neighbors (ANN) model trained to identify the most similar fingerprint based on a subject embedding. For image matching, a score is provided that indicates a similarity between the input embedding and the most similar identified embedding, which allows for matching even when an image has been distorted, rotated, cropped, or otherwise modified. For image classification, the embeddings in the index are clustered and the clusters are labeled. Users can provide just a few images to add to the index as a labeled cluster. The ANN model returns a score and label of the most similar identified embedding for labeling the subject image if the score exceeds a threshold. As improvements are made to the image fingerprinting model, a converter model is trained to convert the original embeddings to be compatible with the new embeddings.

公开(公告)号：US11983955B1

公开(公告)日：2024-05-14

申请号：US18450999

申请日：2023-08-16

Applicant: Netskope, Inc.

Inventor： Emanoel Daryoush ,  Jason B. Bryslawskyj ,  Yi Zhang ,  Ari Azarafrooz ,  Wayne Xin ,  Yihua Liao ,  Niranjan Koduri

IPC: G06V40/12 ,  G06F3/04842 ,  G06T3/60 ,  G06T5/70 ,  G06V10/56

CPC classification number: G06V40/1365 ,  G06F3/04842 ,  G06T3/60 ,  G06T5/70 ,  G06V10/56 ,  G06T2207/10024 ,  G06T2207/20132

Abstract: Image fingerprints (embeddings) are generated by an image fingerprinting model and indexed with an approximate nearest neighbors (ANN) model trained to identify the most similar fingerprint based on a subject embedding. For image matching, a score is provided that indicates a similarity between the input embedding and the most similar identified embedding, which allows for matching even when an image has been distorted, rotated, cropped, or otherwise modified. For image classification, the embeddings in the index are clustered and the clusters are labeled. Users can provide just a few images to add to the index as a labeled cluster. The ANN model returns a score and label of the most similar identified embedding for labeling the subject image if the score exceeds a threshold. As improvements are made to the image fingerprinting model, a converter model is trained to convert the original embeddings to be compatible with the new embeddings.

公开(公告)号：US20250061690A1

公开(公告)日：2025-02-20

申请号：US18451010

申请日：2023-08-16

Applicant: Netskope, Inc.

Inventor： Yihua Liao ,  Niranjan Koduri ,  Emanoel Daryoush ,  Jason B. Bryslawskyj ,  Yi Zhang ,  Ari Azarafrooz ,  Wayne Xin

IPC: G06V10/774 ,  G06F21/62 ,  G06T1/00 ,  G06V10/74 ,  G06V10/762 ,  G06V10/764 ,  G06V10/94 ,  G06V20/70

Abstract: Image fingerprints (embeddings) are generated by an image fingerprinting model and indexed with an approximate nearest neighbors (ANN) model trained to identify the most similar fingerprint based on a subject embedding. For image matching, a score is provided that indicates a similarity between the input embedding and the most similar identified embedding, which allows for matching even when an image has been distorted, rotated, cropped, or otherwise modified. For image classification, the embeddings in the index are clustered and the clusters are labeled. Users can provide just a few images to add to the index as a labeled cluster. The ANN model returns a score and label of the most similar identified embedding for labeling the subject image if the score exceeds a threshold. As improvements are made to the image fingerprinting model, a converter model is trained to convert the original embeddings to be compatible with the new embeddings.