Zeyu Li Phones & Addresses
-
San Diego, CA
-
Pleasanton, CA
-
Livermore, CA
Publications
Us Patents
Position Relationships Associated With Image Capturing Devices
View pageUS Patent:
8300083, Oct 30, 2012
Filed:
Jul 20, 2007
Appl. No.:
11/880396
Inventors:
Henry Harlyn Baker - Palo Alto CA, US
Zeyu Li - Palo Alto CA, US
Robert Samuel Schreiber - Palo Alto CA, US
Zeyu Li - Palo Alto CA, US
Robert Samuel Schreiber - Palo Alto CA, US
Assignee:
Hewlett-Packard Development Company, L.P. - Houston TX
International Classification:
H04N 7/00
US Classification:
348 36, 348 39
Abstract:
A method for determining a plurality of spatial relationships associated with a plurality of image capturing devices is disclosed. In one embodiment, the present method acquires sets of a plurality of source images from a plurality of image capturing devices. The present method then determines a plurality of transforms for each such set of source images for combining the acquired source images into a plurality of seamless images. The present method then determines a plurality of relative positions associated with the plurality of image capturing devices based on the plurality of transforms The present method then determines a plurality of spatial relationships associated with the plurality of image capturing devices based on the transforms and plurality of relative positions associated with the plurality of image capturing devices.
Constructing Arbitrary-Plane And Multi-Arbitrary-Plane Mosaic Composite Images From A Multi-Imager
View pageUS Patent:
20080101724, May 1, 2008
Filed:
Oct 31, 2006
Appl. No.:
11/591432
Inventors:
Henry Harlyn Baker - Palo Alto CA, US
Zeyu Li - Albany CA, US
Zeyu Li - Albany CA, US
International Classification:
G06K 9/00
G06K 9/36
G06K 9/36
US Classification:
382284, 382154
Abstract:
A method for constructing arbitrary-plane and multi-arbitrary-plane mosaic composite images from a multi-imager is disclosed. A first homography set is generated for a multi-imager based on a first reference plane. A second homography set is then generated for the multi-imager based on a second reference plane, where the second reference plane is distinct from the first reference plane. The first homography set and the second homography set are then utilized to generate a third homography set for the multi-imager based on an arbitrary plane. In so doing, the arbitrary plane becomes repositionable with respect to a scene that the multi-imager is viewing. Multiple such arbitrary planes may be selected in constructing multi-planar mosaic composite images. Multiple such second reference planes may be presented to provide improved quality derivation of the said third homography set.
Method And System For Detecting Fraudulent Transactions Using A Fraud Detection Model Trained Based On Dynamic Time Segments
View pageUS Patent:
20220327544, Oct 13, 2022
Filed:
Jun 28, 2022
Appl. No.:
17/852111
Inventors:
- Mountain View CA, US
Runhua ZHAO - San Jose CA, US
Damayanti SENGUPTA - Seattle WA, US
Nicholas John STANG - Minneapolis MN, US
Zeyu LI - Los Angeles CA, US
Runhua ZHAO - San Jose CA, US
Damayanti SENGUPTA - Seattle WA, US
Nicholas John STANG - Minneapolis MN, US
Zeyu LI - Los Angeles CA, US
International Classification:
G06Q 20/40
G06N 3/04
G06N 20/00
G06N 3/04
G06N 20/00
Abstract:
Certain aspects of the present disclosure provide techniques for detecting fraudulent transactions in a transaction processing system. An example method generally includes receiving a request to process a transaction. An input data set including a vector representing the transaction and a plurality of vectors representing historical transactions is generated. The input data set is divided into a plurality of ragged tensors corresponding to non-overlapping time segments of variable length and having a plurality of vectors associated with dates within each time segment A reduced input data set is generated by generating, for each respective ragged tensor of the plurality of ragged tensors, a respective representative vector using max pooling over vectors in the ragged tensor. A fraudulent transaction score is generated based on the reduced input data set using a fraud detection model. The transaction is processed based, at least in part, on the fraudulent transaction score.
Method And System For Detecting Fraudulent Transactions Using A Fraud Detection Model Trained Based On Dynamic Time Segments
View pageUS Patent:
20210312455, Oct 7, 2021
Filed:
Apr 7, 2020
Appl. No.:
16/841967
Inventors:
- Mountain View CA, US
Runhua ZHAO - San Jose CA, US
Damayanti SENGUPTA - Seattle WA, US
Nicholas John STANG - Minneapolis MN, US
Zeyu LI - Los Angeles CA, US
Runhua ZHAO - San Jose CA, US
Damayanti SENGUPTA - Seattle WA, US
Nicholas John STANG - Minneapolis MN, US
Zeyu LI - Los Angeles CA, US
International Classification:
G06Q 20/40
G06N 20/00
G06N 3/04
G06N 20/00
G06N 3/04
Abstract:
Certain aspects of the present disclosure provide techniques for detecting fraudulent transactions in a transaction processing system. An example method generally includes receiving a request to process a transaction. An input data set including a vector representing the transaction and a plurality of vectors representing historical transactions is generated. The input data set is divided into a plurality of ragged tensors corresponding to non-overlapping time segments of variable length and having a plurality of vectors associated with dates within each time segment A reduced input data set is generated by generating, for each respective ragged tensor of the plurality of ragged tensors, a respective representative vector using max pooling over vectors in the ragged tensor. A fraudulent transaction score is generated based on the reduced input data set using a fraud detection model. The transaction is processed based, at least in part, on the fraudulent transaction score.