Microsoft Technology Licensing, LLC

Beijing Jingdong Century Trading Co., Ltd.

Beijing Jingdong Shangke Information Technology Co., Ltd.

Beijing Wodong Tianjun Information Technology Co., Ltd.

Tao Mei

Shipeng Li

Xian-Sheng Hua

Ting Yao

Wu Liu

Ying-Qing Xu

Yong Rui

Linjun Yang

Bowen Zhou

Hailin Shi

Zheng Chen

Bo Zhang

Peng Xu

Jian-Tao Sun

Ning Zhang

He Zhao

Jianlong Fu

Yingwei Pan

Jinfeng Zhuang

Jinlian Guo

Yan Wang

Yue Chen

Jingdong Wang

Yuan Liu

Yang Wang

Wenjun Zeng

Jiawei Liu

Yan-Feng Sun

Kuiyuan Yang

Meng Wang

Chun-Che Wu

Yinglu Liu

Linfang Wang

Fei Sheng

Zhengjun Zha

Zheng-Jun Zha

Teng Li

Yong Wang

Shu Gong

Xingyu Liao

Boqiang Xu

Lingxiao He

Jiawei Zuo

Yehao Li

Kecheng Zheng

Lin-Xie Tang

Qian Bao

Suman Deb Roy

Zhiwei Gu

A method and an electronic device for pedestrian re-identification are provided. The method may include: collecting a target image set including at least two target images, where each target image includes at least one person; extracting a global feature and a head-shoulder feature of each person in each target image in the target image set, where the global feature is an overall appearance feature, and the head-shoulder feature is a feature of a head part and a shoulder part determining a representation feature of each person in each target image based on the global feature and the head-shoulder feature of each person in each target image; and determining a given person in different target images based on the representation feature of each person in each target image.

Pedestrian re-identification method and device

A facial parsing method and apparatus, a facial parsing network training method and apparatus, an electronic device and a non-transitory computer-readable storage medium, which relate to the field of artificial intelligence. The facial parsing method includes inputting a facial image into a pre-trained facial parsing neural network; extracting a semantic feature of the facial image by using a semantic perception sub-network; extracting a boundary feature of the facial image by using a boundary perception sub-network; and processing the cascaded semantic feature and boundary feature by using a fusion sub-network, to obtain a facial region to which each pixel in the facial image belongs. The method can improve the resolution capability of a neural network for boundary pixels between different facial regions of a facial image, thereby improving the precision of facial parsing.

Face parsing method and related devices

A method and an apparatus for retrieving a target are provided. The method may include: obtaining at least one image and a description text of a designated object; extracting image features of the image and text features of the description text by using a pre-trained cross-media feature extraction network; and matching the image features with the text features to determine an image that contains the designated object.

Method and apparatus for retrieving target

An image paragraph description generating method and apparatus, a medium and an electronic device. The method comprises: obtaining image features of an image (S); determining the topic of the image according to the image features by using a convolutional automatic coding method (S); and determining image description information of the image according to the topic by using a long short-term memory (LSTM)-based paragraph coding method (S), wherein the LSTM comprises a sentence-level LSTM and a paragraph-level LSTM.

Image paragraph description generating method and apparatus, medium and electronic device

The present disclosure relates to a broadcast directing method, apparatus and system, and a non-transitory computer-storable medium, which relate to the technical field of computers. The broadcast directing method includes acquiring a reference video stream from a reference camera position; performing event recognition on the reference video stream to obtain at least one reference event frame interval; determining a local sequence of each reference event frame interval according to a correspondence relationship between the event and a camera position identification, the local sequence comprising a camera position identification of each frame image of a video to be played corresponding to the reference event frame interval and a frame identification corresponding to the camera position identification; generating a broadcast directing sequence according to the local sequence; and generating a broadcast directing video according to the broadcast directing sequence and a video stream of a camera position.

Broadcast directing method, apparatus and system

Embodiments of the present application disclose a method and apparatus for detecting a body. A particular embodiment of the method comprises: acquiring a set of candidate body image region in a target image; for a candidate body image region in the set of candidate body image region: acquiring position information and confidences of candidate body key points in the candidate body image region; determining the candidate body key points within a body contour according to body contour information in the candidate body image region and the acquired position information; and determining a confidence score of the candidate body image region according to a sum of the confidences of the candidate body key points within the body contour; and determining a body image region from the set of candidate body image regions according to the confidence scores of the candidate body image regions in set of the candidate body image regions.

Method and apparatus for detecting body

The present disclosure relates to the technical field of image processing, and in particular to an image description generation method, apparatus and system, and a medium and an electronic device. The method comprises: acquiring one or more image region features in a target image, and obtaining a current input vector by performing a mean pooling on the image region features; obtaining respective outer product vectors of the image region features by respectively linearly fusing the current input vector and each of the image region features; calculating, based on the respective outer product vectors of the image region features, an attention distribution of the image region features in a spatial dimension and an attention distribution of the image region features in a channel dimension; and generating an image description of the target image based on the attention distribution of the image region features in the spatial dimension and the attention distribution of the image region features in the channel dimension.

Image description generation method, apparatus and system, and medium and electronic device

Disclosed are a method and device for recognizing a video. One specific embodiment of the method comprises: obtaining a video to be identified; inputting said video to a pre-trained local and global representation propagation LGD model to obtain the category of said video, wherein the LGD model learns a spatial-temporal representation in said video based on diffusion between local and global representations. According to this embodiment, the spatial-temporal representation in the video is learned based on diffusion between the local and global representations.

Method and device for identifying video

The present disclosure provides a self-service settlement method, apparatus and storage medium, and relates to the technical field of self-service shopping, wherein the method includes: obtaining a monitoring image acquired by an image capture device and corresponding to commodities to be settled which are placed on a settlement counter; obtaining information of the commodities to be settled through image recognition; obtaining a first weight acquired by a weight detection device, obtaining a second weight based on the information of the commodities to be settled, judging whether the information of the commodities to be settled is matched with commodities to be confirmed according to a weight comparison result; and obtaining purchased commodity settlement information under the condition that the information of the commodities to be settled is matched with the commodities to be confirmed, and performing checkout processing.

Self-service settlement method, apparatus and storage medium

The present disclosure relates to an information sending method, apparatus, system and a computer readable storage medium, and pertains to the technical field of computers. The method in the present disclosure includes: analyzing video data of an offline user to determine an attribute of the offline user; searching for historical access information of at least one online user matching the attribute of the offline user; and determining at least one object recommended to the offline user according to the historical access information of each of the at least one online user, and sending information of the at least one object to the offline user.

Information sending method, apparatus and system, and computer-readable storage medium

In accordance with implementations of the subject matter described herein, a solution for fine-grained image recognition is proposed. This solution includes extracting a global feature of an image using a first sub-network of a first learning network; determining a first attention region of the image based on the global feature using a second sub-network of the first learning network, the first attention region including a discriminative portion of an object in the image; extracting a first local feature of the first attention region using a first sub-network of a second learning network; and determining a category of the object in the image based at least in part on the first local feature. Through this solution, it is possible to localize an image region at a finer scale accurately such that a local feature at a fine scale can be obtained for object recognition.

Fine-grained image recognition

Implementations of the subject matter described herein provide a solution for extracting spatial-temporal feature representation. In this solution, an input comprising a plurality of images is received at a first layer of a learning network. First features that characterize spatial presentation of the images are extracted from the input in a spatial dimension using a first unit of the first layer. Based on a type of a connection between the first unit and a second unit of the first layer, second features at least characterizing temporal changes across the images are extracted from the first features and/or the input in a temporal dimension using the second unit. A spatial-temporal feature representation of the images is generated partially based on the second features. Through this solution, it is possible to reduce learning network sizes, improve training and use efficiency of learning networks, and obtain accurate spatial-temporal feature representations.

Extraction of spatial-temporal feature representation

Techniques and constructs to facilitate suggestion of image-based search queries can provide personalized trending image search queries. The constructs may enable identification of trending image searches and further personalize those trending image search queries for an identified user based on information about on the user's search history and the search histories of other users. The constructs also may select a representative image for display to the user, such that selection of the representative image will execute the search query. The representative image may be selected from a plurality of candidate images based on its burstiness.

Personalized trending image search suggestion

A facility for using a mobile device to search video content takes advantage of computing capacity on the mobile device to capture input through a camera and/or a microphone, extract an audio-video signature of the input in real time, and to perform progressive search. By extracting a joint audio-video signature from the input in real time as the input is received and sending the signature to the cloud to search similar video content through the layered audio-video indexing, the facility can provide progressive results of candidate videos for progressive signature captures.

Mobile video search

Video description generation using neural network training based on relevance and coherence is described. In some examples, long short-term memory with visual-semantic embedding (LSTM-E) can maximize the probability of generating the next word given previous words and visual content and can create a visual-semantic embedding space for enforcing the relationship between the semantics of an entire sentence and visual content. LSTM-E can include a 2-D and/or 3-D deep convolutional neural networks for learning powerful video representation, a deep recurrent neural network for generating sentences, and a joint embedding model for exploring the relationships between visual content and sentence semantics.

Jointly modeling embedding and translation to bridge video and language

Growing community of inventors

Beijing, China

Tao Mei