Microsoft Technology Licensing, LLC

Twitter, Inc.

Qualtrics, LLC

Northernvue Corporation

Lei Zhang

Oron Nir

Avner Levi

Yuxiao Hu

Irit Ofer

Haim Sabo

Ika Bar-Menachem

Eylon Ami

Maria Zontak

Ella Ben Tov

Tucker Cunningham Burns

Apar Singhal

Xu Yang

Christopher J Buehler

Changhu Wang

Wei Zhang

Hao Liu

Michael James Miles

Yandong Guo

Cornelia Carapcea

Sheng Zhang

Shanshan Liu

Jian Zhou

Yucao Wang

Ao Guo

Miao Zou

Doo Suk Yoon

Tianjing Xu

Ameet Ranadive

Shariq Rizvi

Dongdong Chen

Sandeep Pandey

Chinmoy Dutta

Lu Yuan

Avneesh Sud

Santosh Kumar Kancha

David Benjamin Liu

Xiaochuan Qin

Abhishek Shrivastava

Yimin Wu

Ke Zhou

Junjun Li

Wanchen Lu

Vibhor Rastogi

Milind Mahajan

Danmeng Liu

Anika Zaman

Xinghai Sun

Jie Jerry Liu

Nissanka Arachchige Bodhi Priyantha

Di Wang

Suman Kumar Nath

Dimitrios Lymberopoulos

Bin Wang

Gerald Reuben DeJean

Robert Zhu

Ying Chin

Yinpeng Chen

Pengxiang Zhao

Michel Goraczko

Dejun Zhang

David Gregory Grant

Mohammed Shoaib

Satyendra Bahadur

Myung Ryul Jang

Mengchen Liu

Xiyang Dai

Marcel Gavriliu

Daryl A Welsh

Carlos Picoto

Weijia Mao

Anna Roth

Minglun Qian

Qiankun Liu

Bin Xiao

Sean Bergam

Weijun Guo

Chittibabu Pacharu

Paul Martin Messing

Xiaowen Wang

Kevin Alan Erickson

Shenghao Li

Ying N Chin

David Liu

Robert Yu Zhu

Zhiwei Li

A system for tracking a target object across a plurality of image frames. The system comprises a logic machine and a storage machine. The storage machine holds instructions executable by the logic machine to calculate a trajectory for the target object over one or more previous frames occurring before a target frame. Responsive to assessing no detection of the target object in the target frame, the instructions are executable to predict an estimated region for the target object based on the trajectory, predict an occlusion center based on a set of candidate occluding locations for a set of other objects within a threshold distance of the estimated region, each location of the set of candidate occluding locations overlapping with the estimated region, and automatically estimate a bounding box for the target object in the target frame based on the occlusion center.

Occlusion-aware multi-object tracking

The present disclosure relates to systems, non-transitory computer-readable media, and methods for applying attribute definitions to various survey data and utilizing attribute definitions to generate survey analyses. In particular, in one or more embodiments, the disclosed systems intelligently determines attribute definitions to apply to digital surveys, digital survey questions, and digital survey responses. The disclosed systems can determine attribute definitions to apply to survey data based on user input, associated survey templates, associated prior use of attribute definition, analysis of the text of digital surveys, and a variety of other attributes of survey data. The disclosed systems can generate digital survey analysis by utilizing attribute definitions applied to a variety of types of survey data.

Determining and applying attribute definitions to digital survey data to generate survey analyses

Some disclosed embodiments are directed to methods and systems for performing pair-wise delta compression. For example, systems obtain a set of files to be compressed into a single compressed file. The system identifies different attributes related to the set of files. For each file in the set of files, the system predicts an optimized set of candidate compression files and calculates a delta between each file in the optimized set and the target file corresponding to the optimized set. After identifying the smallest delta, the system compresses the selected pair of files associated with the smallest delta in order to generate the single compressed file for the set of files.

Systems and methods for pair-wise delta compression

Systems and methods for object detection generate a feature pyramid corresponding to image data, and rescaling the feature pyramid to a scale corresponding to a median level of the feature pyramid, wherein the rescaled feature pyramid is a four-dimensional (4D) tensor. The 4D tensor is reshaped into a three-dimensional (3D) tensor having individual perspectives including scale features, spatial features, and task features corresponding to different dimensions of the 3D tensor. The 3D tensor is used with a plurality of attention layers to update a plurality of feature maps associated with the image data. Object detection is performed on the image data using the updated plurality of feature maps.

Dynamic head for object detection

The technology described herein is directed to systems, methods, and software for indexing video. In an implementation, a method comprises identifying one or more regions of interest around target content in a frame of the video. Further, the method includes identifying, in a portion of the frame outside a region of interest, potentially empty regions adjacent to the region of interest. The method continues with identifying at least one empty region of the potentially empty regions that satisfies one or more criteria and classifying at least the one empty region as a negative sample of the target content. In some implementations, the negative sample of the target content in a set of negative samples of the target content, with which to train a machine learning model employed to identify instances of the target content.

Negative sampling algorithm for enhanced image classification

A self-contained game monitoring device captures images of a game table and identifies objects relevant to the game and identifies values associated with objects. During the course of the game, the device can detect a violation of the players' bets according to pre-configured game rules. At the end of each game, the device can determine the outcome of the game (e.g., the win/lose/push on each bet) and can determine whether the dealer's action (e.g., payout on each bet) is consistent with the device's judgment. If an inconsistent action is detected, the device can notify the dealer/supervisor about a potential mistake.

Game monitoring device

A method includes obtaining training data for a classifier, the training data comprises one or more target classes, obtaining candidate background classes, selecting negative classes from the candidate background classes, wherein the negative classes exclude candidate background classes that are close to the target classes, wherein the negative classes exclude candidate background classes that are very different from the target classes, and wherein the negative classes include candidate background classes that are similar to the target classes, and training the classifier on a combined set of the selected negative classes and target classes.

Adding negative classes for training classifier

The technology described herein is directed to a media indexer framework including a character recognition engine that automatically detects and groups instances (or occurrences) of characters in a multi-frame animated media file. More specifically, the character recognition engine automatically detects and groups the instances (or occurrences) of the characters in the multi-frame animated media file such that each group contains images associated with a single character. The character groups are then labeled and used to train an image classification model. Once trained, the image classification model can be applied to subsequent multi-frame animated media files to automatically classifying the animated characters included therein.

Semi supervised target recognition in video

The present disclosure provides technical solutions related to intelligent cloud management based on profile. Artificial intelligent is applied to cloud management and cloud management suggestion may be proposed intelligently. In daily work, behaviors in using cloud resources may show characteristics of cloud users or cloud tenants themselves. The technical solution of intelligent cloud management of the present disclosure generates profile identifying cloud using characteristics by extracting behavior data in using cloud and intelligently proposes cloud management suggestions based on the profile.

Intelligent cloud management based on profile

The present disclosure provides technical solutions related to action undo service based on cloud platform. Related operations dependent on target operations to be undone may be obtained by dependency analysis and the target operations and the related operations may be undone by time sequence so as to reduce the conflict caused by the undo operations.

Action undo service based on cloud platform

Intelligent assistant for cloud management

Semi supervised animated character recognition in video

A method for conversion attribution. The method includes obtaining a first identifier associated with a first device, obtaining a second identifier associated with a second device, bridging the first identifier and the second identifier based on a determination, made by a probabilistic classifier, that the first identifier and the second identifier are associated with a common user, and attributing, using the bridge, a conversion on a website accessed using the second device.

Method and system for identifying users across mobile and desktop devices

Methods, systems, and computer programs are presented for training a deep neural network (DNN). One method includes an operation for training a predecessor network defined for image recognition of items, where parameters of a predecessor classifier are initialized with random numbers sampled from a predetermined distribution, and the predecessor classifier utilizes an image-classification probability function without bias. The method further includes an operation for training a successor network defined for image recognition of items in a plurality of classes, where parameters of a successor classifier are initialized with parameters learned from the predecessor network, and the successor classifier utilizes the image-classification probability function without bias. Further, the method includes operations for receiving an image for recognition, and recognizing the image utilizing the successor classifier.

Fast deep neural network training

Growing community of inventors

Bellevue, WA, United States of America

Lei Zhang