Microsoft Technology Licensing, LLC

Beijing University of Civil Engineering and Architecture

Fang Wen

Jian Qiao Sun

Xiaoou Tang

Ying-Qing Xu

Dong Chen

Gang Hua

Lin Liang

Yichen Wei

Heung-Yeung Shum

Jinyu Li

Xin Tong

Rong Xiao

Chuan Liu

Qiong Yang

Jianwei Yang

Ran Sun

Tangbo Bai

Ziming Kou

Jianmin Bao

Zhifen Pang

Dechen Yao

Tommer Leyvand

Qifa Ke

Michael John Conrad

Casey Leon Meekhof

Yi Li

Lu Yuan

Yu-ting Kuo

David P Vronay

Chun-Te Chu

Dongdong Chen

Yue Ma

Tim Keosababian

Justin Hamilton

Xin Zou

Xiao-Ou Tang

Hsiang-Tao Wu

Yangang Wang

Mizuki Oka

Baining Guo

Hao Yang

Ting Zhang

Lingzhi Li

Jiaolong Yang

Xiaoyi Dong

Aamir Jawaid

The disclosure herein describes training an encoder network to inpaint images with masked portions. A primary encoding process is used to encode a visible portion of a masked input image into encoded token data. The encoded token data is then decoded into both pixel regression output and feature prediction output, wherein both outputs include inpainted image data associated with the masked portion of the masked input image. A pixel regression loss is determined using the pixel regression output and pixel data of an unmasked version of the masked input image. A feature prediction loss is determined using the feature prediction output and ground truth encoding output of the unmasked version of the masked input image. The primary encoding process is then trained using the pixel regression loss and the feature prediction loss, whereby the primary encoding process is trained to encode structural features of input images into encoded token data.

Training masked autoencoders for image inpainting

In implementations of the subject matter as described herein, there is provided a method for forgery detection of a face image. Subsequent to inputting a face image, it is detected whether a blending boundary due to the blend of different images exists in the face image, and then a corresponding grayscale image is generated based on a result of the detection, where the generated grayscale image can reveal whether the input face image is formed by blending different images. If a visible boundary corresponding to the blending boundary exists in the generated grayscale image, it indicates that the face image is a forged image; on the contrary, if the visible boundary does not exist in the generated grayscale image, it indicates that the face image is a real image.

Forgery detection of face image

In accordance with implementations of the present disclosure, there is provided a solution for portrait editing and synthesis. In this solution, a first image about a head of a user is obtained. A three-dimensional head model representing the head of the user is generated based on the first image. In response to receiving a command of changing a head feature of the user, the three-dimensional head model is transformed to reflect the changed head feature. A second image about the head of the user is generated based on the transformed three-dimensional head model, and reflects the changed head feature of the user. In this way, the solution can realize editing of features like a head pose and/or a facial expression based on a single portrait image without manual intervention and automatically synthesize a corresponding image.

Portrait editing and synthesis

In implementations of the subject matter described herein, a solution for object detection is proposed. First, a feature(s) is extracted from an image and used to identify a candidate object region in the image. Then another feature(s) is extracted from the identified candidate object region. Based on the features extracted in these two stages, a target object region in the image and a confidence for the target object region are determined. In this way, the features that characterize the image from the whole scale and a local scale are both taken into consideration in object recognition, thereby improving accuracy of the object detection.

Object detection based on joint feature extraction

In accordance with implementations of the subject matter described herein, there is provided a solution for face synthesis. In this solution, a first image about a face of a first user and a second image about a face of a second user are obtained. A first feature characterizing an identity of the first user is extracted from the first image, and a second feature characterizing a plurality of attributes of the second image is extracted from the second image, where the plurality of attributes do not include the identity of the second user. Then, a third image about a face of the first user is generated based on the first and second features, the third image reflecting the identity of the first user and the plurality of attributes of the second image.

Face synthesis

An intelligent storage control system is provided in the present invention. The system includes a tower body, a delivery module, a drive module, a clamping module, an operation module, a control module, and a communication unit. When the control module receives a storage command transmitted by the operation module, the main controller is configured to analyze whether there is free storage space. If there is a free storage space, the control module controls the drive module to drive the delivery module to push the bicycle to the storage space of the tower body, and a serial number of the storage space is displaced on a display screen. The present invention can operate conveniently, and can save the footprint area of bicycles.

Intelligent storage control system

A three-dimensional rotating intelligent storage compartment is provided in the present invention, which includes a tower body, a center of the tower body being provided with a central garage for storing bicycles, the central garage including a central pillar. A plurality of layers of placement plates are arranged around the central pillar, and each layer of the placement plates is equally divided into a plurality of independent storage spaces by a longitudinal partitioning plate. A longitudinal edge on one side of the longitudinal partitioning plate is fixed to the central pillar; each storage space is provided with a clamping mechanism for clamping bicycle. An inner wall of the tower body is longitudinally provided with a plurality of loop tracks, each loop track is provided with a plurality of ferry parking spaces, which are movable within the loop track, is capable of clamping bicycle in a standing position and capable of pushing bicycle to the placement plate. The present invention can operate conveniently, and can save footprint area of bicycles.

Three-dimensional rotating intelligent storage compartment

The present disclosure provides method, apparatus and system for 3-dimension (3D) face tracking. The method for 3D face tracking may comprise: obtaining a 2-dimension (2D) face image; performing a local feature regression on the 2D face image to determine 3D face representation parameters corresponding to the 2D face image; and generating a 3D facial mesh and corresponding 2D facial landmarks based on the determined 3D face representation parameters. The present disclosure may improve tracking accuracy and reduce memory cost, and accordingly may be effectively applied in broader application scenarios.

Method, apparatus and system for 3D face tracking

Examples are disclosed herein that relate to detecting spoofed human faces. One example provides a computing device comprising a processor configured to compute a first feature distance between registered image data of a human face in a first spectral region and test image data of the human face in the first spectral region, compute a second feature distance between the registered image data and test image data of the human face in a second spectral region, compute a test feature distance between the test image data in the first spectral region and the test image data in the second spectral region, determine, based on a predetermined relationship, whether the human face to which the test image data in the first and second spectral regions corresponds is a real human face or a spoofed human face, and modify a behavior of the computing device.

Spoofed face detection

Disclosed herein are techniques and systems for computing geodesic saliency of images using background priors. An input image may be segmented into a plurality of patches, and a graph associated with the image may be generated, the graph comprising nodes and edges. The nodes of the graph include nodes that correspond to the plurality of patches of the image plus an additional virtual background node that is added to the graph. The graph further includes edges that connect the nodes to each other, including internal edges between adjacent patches and boundary edges between those patches at the boundary of the image and the virtual background node. Using this graph, a saliency value, called the 'geodesic' saliency, for each patch of the image is determined as a length of a shortest path from a respective patch to the virtual background node.

Geodesic saliency using background priors

Methods and computer-storage media having computer-executable instructions embodied thereon that facilitate refining query results using visual cues are provided. Query results are determined in response to an indication of a user query. One or more groups of query results are generated from the query results based on categories of query results that share similar features. Visual cues are associated with each of the query result groups. Visual cues, in association with query result groups, are presented to a user. Query results associated with a selected visual cue may be presented to a user. A refined user query may be generated based on a selected visual cue.

Visual-cue refinement of user query results

Embodiments that provide cartoon personalization are disclosed. In accordance with one embodiment, cartoon personalization includes selecting a face image having a pose orientation that substantially matches an original pose orientation of a character in a cartoon image. The method also includes replacing a face of the character in the cartoon image with the face image. The method further includes blending the face image with a remainder of the character in the cartoon image.

Cartoon personalization

A method for creating an optimized gradient mesh of a vector-based image from a raster-based image. In one implementation, a set of boundaries for an object on a raster-based image may be received. An initial gradient mesh of the object may be created. A residual energy between the object on the raster-based image and a rendered initial gradient mesh may be minimized to generate an optimized gradient mesh.

Creating optimized gradient mesh of a vector-based image from a raster-based image

Image search techniques are described. In one or more implementations, images in a search result are ordered based at least in part on similarity of the images, one to another. The search result having the ordered images is provided in response to a search request.

Method for searching and ranking images clustered based upon similar content

Growing community of inventors

Beijing, China

Fang Wen