International Business Machines Corporation

Microsoft Technology Licensing, LLC

Facebook Technologies, LLC

Disney Enterprises, Inc.

Carnegie Mellon University

Takaaki Shiratori

Xin Tong

Xiang Cao

Yang Liu

Nobuyuki Umetani

Yupeng Zhang

Teng Han

Zhimin Ren

Yaser Sheikh

Jason Saragih

Tomas Simon Kreuz

Chenglei Wu

Tomoka Mochizuki

Stephen Anthony Lombardi

Takehiko Ishii

Shoou-I Yu

Sing Bing Kang

Xiaoou Tang

Tommer Leyvand

Leonid Sigal

Feng Xu

Jessica Kate Hodgins

Yasuyuki Matsushita

Akio Kido

Thomas Gersten

Masahiko Maedera

Karl D D Willis

Hyun Soo Park

Shigeru Tachikawa

Tsuneo Mochizuki

In a method for improving machine translation results, a processor receives a target document in a first language. A processor may also translate the target document into a first translated document in a second language using a first neural machine translation (NMT) model, determine a target attribute value from the target document and a first translated attribute value for the first translated document using a natural language understanding (NLU) model, and compare the target attribute value to the first translated attribute value to determine a first comparison score for the first NMT model.

Detection and correction of mis-translation

An embodiment includes determining semantic attributes of a bookmark. The embodiment also includes determining respective attribute values for each of the semantic attributes of the bookmark. The embodiment also includes rendering a three-dimensional (3D) virtual space for display to a user, where the virtual space is defined by three orthogonal axes, each associated with a respective one of the semantic attributes. The embodiment also includes displaying a symbol representative of the bookmark in the virtual space. The symbol is positioned in the virtual space at an intersection of perpendicular projections from locations on the three axes corresponding with respective attribute values of the attributes associated with the respective axes.

Attribute-based positioning of bookmarks in a 3D virtual space

An intuitive interface may allow users of a computing device (e.g., children, etc.) to create imaginary three dimensional (3D) objects of any shape using body gestures performed by the users as a primary or only input. A user may make motions while in front of an imaging device that senses movement of the user. The interface may allow first-person and/or third person interaction during creation of objects, which may map a body of a user to a body of an object presented by a display. In an example process, the user may start by scanning an arbitrary body gesture into an initial shape of an object. Next, the user may perform various gestures using his body, which may result in various edits to the object. After the object is completed, the object may be animated, possibly based on movements of the user.

Object creation using body gestures

In one embodiment, a system may access a codec that encodes an appearance associated with a subject and comprise codec portions that respectively correspond to body parts of the subject. The system may generate a training codec that comprises a first subset of the codec portions (a first set of body parts) and a modified second subset of the codec portions (muted body parts). The system may decode the training codec using a machine-learning model to generate a mesh of the subject. The system may transform the mesh of the subject based on a predetermined pose. The system may update the machine-learning model based on a comparison between the transformed mesh and a target mesh of the subject having the predetermined pose. The system in the present application can train a machine-learning model to render an avatar with a pose using uncorrelated codec portions corresponding to different body parts.

Systems and methods for rendering avatar using uncorrelated codec portions in codec

The disclosed computer system may include an input module, an autoencoder, and a rendering module. The input module may receive geometry information and images of a subject. The geometry information may be indicative of variation in geometry of the subject over time. Each image may be associated with a respective viewpoint and may include a view-dependent texture map of the subject. The autoencoder may jointly encode texture information and the geometry information to provide a latent vector. The autoencoder may infer, using the latent vector, an inferred geometry and an inferred view-dependent texture of the subject for a predicted viewpoint. The rendering module may be configured to render a reconstructed image of the subject for the predicted viewpoint using the inferred geometry and the inferred view-dependent texture. Various other systems and methods are also disclosed.

Systems and methods for rendering avatars with deep appearance models

An optical flow of depth video of a depth camera imaging a human subject is recognized. An energy field created by motion of the human subject is generated as a function of the optical flow and specified rules of a physical simulation of the virtual environment. The energy field is mapped to a virtual position in the virtual environment. A property of a virtual object in the virtual environment is adjusted based on a plurality of energy elements of the energy field in response to the virtual object interacting with the virtual position of the energy field.

Motion control of a virtual environment

Techniques for presenting display frames projected using a handheld projector. Embodiments detect a marker embedded in a display surface. A first display frame is projected onto the display surface from the handheld projector, where the first display frame projects, on the display surface, one or more animated objects positioned relative to the detected marker embedded in the display surface. Embodiments generate one or more subsequent display frames projected from the handheld projector, where at least a first one of the animated objects in the one or more subsequent display frames is positioned relative to the detected marker embedded in the display surface.

Trackable projection surfaces using hidden marker tracking

Body-mounted cameras are used to accurately reconstruct the motion of a subject. Outward-looking cameras are attached to the limbs of the subject, and the joint angles and root pose that define the subject's configuration are estimated through a non-linear optimization, which can incorporate image matching error and temporal continuity of motion. Instrumentation of the environment is not required, allowing for motion capture over extended areas and in outdoor settings.

Motion capture from body mounted cameras

Methods and apparatus, including computer program products, for providing assistance in detecting mistranslation in a translated document obtained by performing machine translation of an original document. A word included in the translated document is obtained. Search results are obtained of searching both a first document data group and a second document data group including the first document data group for pieces of document data related to the obtained word. Based on the obtained search results, an index is generated. The index indicates the adequacy of the obtained word as a translated word in a field corresponding to the first document data group. The generated index is output.

Apparatus and methods for providing assistance in detecting mistranslation

There is a system for selecting a text for output from among multiple texts according to whether or not a phrase is included in the multiple texts. The system includes a determination unit that determines whether or not each of predetermined multiple phrases is included in each of the multiple texts; an index calculation unit that, with respect to each of the multiple phrases, calculates a larger index for a text having the highest of priorities preset among texts including the phrase, then indexes for other texts; and an output unit that selects texts each having a larger sum of sums of the calculated index for each text and outputs a group of the selected texts.

Technology for selecting texts suitable as processing objects

Growing community of inventors

Tokyo, Japan

Takaaki Shiratori