International Business Machines Corporation

Nuance Communications, Inc.

Globalfoundries U.S. 2 LLC

Shi Lei Zhang

Yong Qin

Qin Shi

Wen Liu

Zhong Su

Jian Ye Chen

Qing Fa Wang

Feng Jin

Yi Qin Yu

Jing Li

Guo Tong Xie

Haifeng Liu

Zhiwei Shuang

Xin Zhang

Jie Zhang

Peng Wang

Peng Gao

Sheng Hua Bao

Shiwan Zhao

Liu Wen

Yubo Li

Shenghua Bao

Xiaolu Zhang

Ke Jin

Wei Shan Dong

Min Li

Yonghua Lin

Shi Wan Zhao

Junsong Wang

Lingyun Wang

Eryu Xia

Bibo Hao

Xiu Li Li

Songfang Huang

Qin Jin

Xu Dong Tu

Su Liu

Wei Sun

Tao Wang

Jian Ding

Rong Yan

Ning Duan

Cheng Xu

Jing Min Xu

Lin Y Yang

Fan Jing Meng

Xiao Xun Zhang

Yi Liu

Jun Chi Yan

Jie Zhou

Li Zhang

Qing Bo Wang

Peng Fei Zhu

Ke Wei Sun

Naga A Ayachitula

Ryo Kawahara

En Liang Xu

Keith C Houck

Zhi Wei Shuang

Fan Ping Meng

Li Jun Ma

Toyotaro Suzumura

Zhuo Su

Bingzhe Wu

Peng Cheng P P Zhu

Shouyu Yan

Sheng Hu Bao

A suspicious pattern can be detected within a transaction network of nodes. Nodes of the network are walked by determining if any adjacent node to a current node is within the suspicious pattern. If an adjacent node is within the suspicious pattern, that node is walked to. Based on the walk, a node sequence can be generated.

Generating sequences of network nodes

A computer-implemented method for data augmentation is provided according an embodiment of the present disclosure. In the method, a first feature vector for input data may be obtained based on a first model. The input data may be clustered to a plurality of clusters. For each of the clusters, a second feature vector may be obtained based on the first model. Then, a similarity between the first feature vector and the second feature vector may be estimated for each of the clusters. At least one cluster of the plurality of clusters for which the similarity is lower than a threshold may be determined. Moreover, data augmentation may be performed to the at least one cluster.

Data augmentation

Embodiments of the invention are directed to a computer-implemented method that includes using a reinforcement learning (RL) system to generate a first set of displayed region candidates based on inputs received from online users while watching video. A recommendation system is used to rank the first set of displayed region candidates based on inputs received from a local user watching video. The recommendation system is further used to select a first highest ranked one of the first set of displayed region candidates. Based on the first highest ranked one of the first set of displayed region candidates, a first section of a first raw video frame is fetched that matches the first highest ranked one of the first set of displayed candidate regions, wherein the first section of the first raw video frame includes a first predicted display region of the video frame.

Using reinforcement learning and personalized recommendations to generate a video stream having a predicted, personalized, and enhance-quality field-of-view

Traffic jam patterns can be identified and, based on historical traffic data, pre-traffic jam patterns that are likely to result in the traffic jam patterns can be identified as well. Real-time traffic data regarding a driving road of a community can be received and analyzed to determine whether the real-time traffic data match with a pre-traffic jam pattern. If the data matches a pre-traffic jam pattern, an alerting signal for predicting a traffic jam can be transmitted.

Traffic data analysis and traffic jam prediction

A method can include obtaining, from a video having a first resolution, a set of frames having a second resolution. The first resolution can be higher than the second resolution. The set of frames can include a first frame and a second frame. The method can include generating a difference feature map. The method can include obtaining a third frame having the first resolution. The method can include detecting, based on the difference feature map, a first location of a first object in the third frame. The method can include cropping, from the third frame, a first cropped area. The first cropped area can be smaller than a third frame area. The method can include generating, based on a feature map and the difference feature map, a spatial attention layer. The method can include detecting, by the spatial attention layer, the first object in the first cropped area.

Difference-guided video analysis

Techniques for candidate evaluation and filtering are provided. Enrollment criteria for a clinical trial are received, where the enrollment criteria include a plurality of conditions. A plurality of cost vectors is constructed for the plurality of conditions. A set of values for one or more of the plurality of conditions is determined for a candidate, where the set of values does not include a value for at least a first condition of the plurality of conditions. A utilized cost is generated for the candidate, based on the first set of values and the plurality of cost vectors. The candidate is then ranked based on the utilized cost.

Processing profiles using machine learning to evaluate candidates

A method for detecting deviations between an event log and a process model includes converting the process model into a probability process model, the probability process model comprising multiple nodes in multiple hierarchies and probability distribution associated with the multiple nodes, a leaf node among the multiple nodes corresponding to an activity in the process model; detecting differences between at least one event sequence contained in the event log and the probability process model according to a correspondence relationship; and identifying the differences as the deviations in response to the differences exceeding a predefined threshold; wherein the correspondence relationship describes a correspondence relationship between an event in one event sequence of the at least one event sequence and a leaf node in the probability process model.

Detecting deviations between event log and process model

Techniques for unattended object detection using machine learning are disclosed. A machine learning policy, for use in identifying unattended objects in a captured image depicting one or more objects, is generated. The generating includes determining a level of occlusion in the captured image relating to the objects, and determining the machine learning policy based on the determined level of occlusion. A machine learning model is selected, from among a plurality of pre-defined machine learning models, based on the generated machine learning policy. An unattended object is detected in the captured image using the selected machine learning model.

Unattended object detection using machine learning

A computer-implemented method for attention generation is provided. In this method, a plurality of image frames can be obtained from a video stream. An original attention for a first image frame of the plurality of image frames can be generated. Then, at least one interested area can be identified in the first image frame. A local attention for each of the at least one interested area can be generated. Moreover, a total attention for the first image frame can be generated based on the original attention of the first image frame and the local attention of each of the at least one interested area.

Attention generation

Embodiments of the present disclosure relate to object detection in an image. In an embodiment, a computer-implemented method is disclosed. According to the method, image data representing a scene is obtained and sound distribution information related to the scene is obtained. A detection strategy to be applied in object detection is determined based on the sound distribution information. The object detection is performed on the image data by applying the detection strategy. In other embodiments, a system and a computer program product are disclosed.

Object detection in an image

Systems and methods for ticket classification and response include labeling tickets with a ticket classifier that assigns a ticket label and an associated confidence score to each ticket. Tickets are clustered according to semantic similarity to form ticket clusters. A template associated with each ticket cluster is determined. Templates and the respective ticket clusters are clustered according to semantic similarity to form one or more ticket super-clusters. Tickets that have below-threshold confidence scores are labeled according to the one or more ticket super-clusters. The tickets are automatically responded to.

Hybrid learning-based ticket classification and response

Aspects of the present invention disclose a method, computer program product, and system for image processing. The method includes one or more processors generating a first set of binary images from a first image based on a first color attribute value range associated with a textual object presented in the first image. The method further includes one or more processors recognizing a first set of candidates for the textual object from the first set of binary images. The method further includes one or more processors determining a first appearance frequency of a first candidate in the first set of candidates. In response to determining that the first appearance frequency exceeds a first frequency, threshold he method further includes one or more processors determining that the first candidate is a first recognition result for the textual object in the first image.

Image recognition

Techniques facilitating attention based sequential image processing are provided. A system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise an initialization component that can perform self-attention based training on a model that comprises context information associated with a sequence of images. Images of the sequence of images can be selected during the self-attention based training. The computer executable components can also comprise a localization component that can extract local information from the images selected during the self-attention based training based on the context information. In addition, the computer executable components can also comprise an integration component that can update the model based on an end-to-end integrated attention training framework comprising the context information and the local information.

Attention based sequential image processing

In some examples, a method includes receiving a test image from a user or client, the test image depicting an anatomical region or structure of a human body, obtaining a reference image corresponding to the anatomical region or structure depicted in the test image, and analyzing the test image and the reference image to obtain a set of differences between the two images. In some examples, the method further includes, based at least in part on the set of differences, detecting a possible abnormality in the test image and outputting a result to the user or client. In some examples, a fast R-CNN is used to detecting the possible abnormality in the test image.

Reference image guided object detection in medical image processing

Growing community of inventors

Beijing, China

Shi Lei Zhang