Sudhir Kumar Singh, 55Bryant, WA

2012010, Apr 26, 2012

Apr 21, 2011

13/091459

Nima Sarshar - Fremont CA, US
Sudhir Kumar Singh - San Jose CA, US
Vwani P. Roychowdhury - Los Angeles CA, US

G06F 17/30

707737, 707E17084, 707E17091, 707E17009, 707E17101, 707E17028

The present disclosure describes a method and system called “Universal Learner (UL),” which provides a unified framework to understand multimedia signals. The UL utilizes the loosely annotated multimedia data on the Web, analyses it in various signal domains, such as text, image, audio and combinations thereof, and builds an association graph called the “Multimedia Brain,” which basically comprises visual signals, audio signals, text phrases and the like that capture a multitude of objects, experiences and their attributes and the links among them that capture similar intent or functional and contextual relationships.

2012014, Jun 7, 2012

Dec 6, 2011

13/312752

Sudhir Kumar Singh - San Jose CA, US
Nima Sarshar - Fremont CA, US
Vwani Roychowdhury - Los Angeles CA, US

G06K 9/00
G06K 9/34

382103, 382164

Methods and systems for discovering styles via color and pattern co-occurrence are disclosed. According to one embodiment, a computer-implemented method comprises collecting a set of fashion images, selecting at least one subset within the set of fashion images, the subset comprising at least one image containing a fashion item, and computing a set of segments by segmenting the at least one image into at least one dress segment. Color and pattern representations of the set of segments are computed by using a color analysis method and a pattern analysis method respectively. A graph is created wherein each graph node corresponds to one of a color representation or a pattern representation computed for the set of segments. Weights of edges between nodes of the graph indicate a degree of how the corresponding colors or patterns complement each other in a fashion sense.

2013007, Mar 21, 2013

Sep 19, 2012

13/622917

Sudhir Kumar Singh - San Jose CA, US
Vwani Roychowdhury - Los Angeles CA, US

G09B 19/00

434 81

This invention presents a universal framework for the discovery, understanding and matching of dress styles. In one embodiment, a computer-implemented method for building a universal dress style learner is disclosed, said method comprising: learning human skin models; detecting skin using the learned human skin models; collecting a set of dress images worn by a model; computing a set of style features based on the skin detected for at least one subset within the set of dress images; computing a set of clusters on the at least one subset of dress images based on at least one subset of the set of style features; validating the set of clusters for the at least one subset of style features; and computing a set of validated style features and a style basis.

2013032, Dec 5, 2013

May 14, 2013

13/894313

Nima Sarshar - Fremont CA, US
Sudhir Kumar Singh - San Jose CA, US
Vwani P. Roychowdhury - Los Angeles CA, US

G06F 17/30

707737

The present disclosure describes a method and system called “Universal Learner (UL),” which provides a unified framework to understand multimedia signals. The UL utilizes the loosely annotated multimedia data on the Web, analyses it in various signal domains, such as text, image, audio and combinations thereof, and builds an association graph called the “Multimedia Brain,” which basically comprises visual signals, audio signals, text phrases and the like that capture a multitude of objects, experiences and their attributes and the links among them that capture similar intent or functional and contextual relationships.

2020020, Jun 25, 2020

May 31, 2018

16/613365

- Santa Clara CA, US
Sudhir K. Singh - Dublin CA, US
Vinod Sharma - Menlo Park CA, US
Malini Krishnan Bhandaru - San Jose CA, US

G06N 3/08
G06N 5/04
G06N 20/10

A quaternion deep neural network (QTDNN) includes a plurality of modular hidden layers, each comprising a set of QT computation sublayers, including a quaternion (QT) general matrix multiplication sublayer, a QT non-linear activations sublayer, and a QT sampling sublayer arranged along a forward signal propagation path. Each QT computation sublayer of the set has a plurality of QT computation engines. In each modular hidden layer, a steering sublayer precedes each of the QT computation sublayers along the forward signal propagation path. The steering sublayer directs a forward-propagating quaternion-valued signal to a selected at least one QT computation engine of a next QT computation subsequent sublayer.

2020019, Jun 18, 2020

May 31, 2018

16/613349

- Santa Clara CA, US
Sudhir K. Singh - dublin CA, US
Vinod Sharma - Menlo Park CA, US
Malini Krishnan Bhandaru - San Jose CA, US

G06K 9/62
G06N 3/08
G06N 3/04

A deep neural network (DNN) includes hidden layers arranged along a forward propagation path between an input layer and an output layer. The input layer accepts training data comprising quaternion values, outputs a quaternion-valued signal along the forward path to at least one of the hidden layers. At least some of the hidden layers include quaternion layers to execute consistent quaternion (QT) forward operations based on one or more variable parameters. A loss function engine produces a loss function representing an error between the DNN result and an expected result. QT backpropagation-based training operations include computing layer-wise QT partial derivatives, consistent with an orthogonal basis of quaternion space, of the loss function with respect to a QT conjugate of the one or more variable parameters and of respective inputs to the quaternion layers.

2020011, Apr 16, 2020

May 31, 2018

16/613380

- Santa Clara CA, US
Sudhir K. Singh - Dublin CA, US
Vinod Sharma - Menlo Park CA, US
Malini Krishnan Bhandaru - San Jose CA, US

G06N 3/08
G06N 3/04
G06N 10/00
G06F 17/16
G06K 9/62

A machine-learning system includes a quaternion (QT) computation engine. Input data to the QT computation engine includes quaternion values, each comprising a real component and three imaginary components, represented as a set of real-valued tensors. A single quaternion value is represented as a 1-dimensional real-valued tensor having four real-valued components, wherein a first real-valued component represents the real component of the single quaternion value, and wherein a second, a third, and a fourth real-valued component each respectively represents one of the imaginary components. A quaternion-valued vector having a size N is represented as a 2-dimensional real-valued tensor comprising N 1-dimensional real-valued tensors. A quaternion-valued matrix having N×M dimensions is represented as a 3-dimensional real-valued tensor comprising M 2-dimensional real-valued tensors comprising N 1-dimensional real-valued tensors.

2019025, Aug 22, 2019

Feb 19, 2019

16/279913

- Mountain View CA, US
Sudhir Kumar SINGH - Mountain View CA, US

INVII.AI - Mountain View CA

G10L 15/22
G10L 17/00
G06N 3/02
G06K 9/00

Methods and systems for building a universal always-on multimodal identification system. A universal representation to be used for executing one or more tasks, working on data with one or more signal modalities and comprising modal fusions signals at various levels is learned from a dataset that is targeted user or object agnostic. This universal representation is combined with a second stage task specific representation that is learned on-the-device using data from the particular user without sending the data to the cloud. The universal representation in combination with the downstream task specific representation is used to build a system to identify people and things using their visual appearances as well as voice by combining scores from one, two or more of the tasks such as face recognition and text independent voice recognition, wherein all required computation for the identification is performed completely on-the-device and no raw data from the user is sent to the cloud without explicit permission of an authorized user.