Nvidia Corporation

Samsung Electronics Co., Ltd.

Kihwan Kim

Jan Kautz

Shalini De Mello

Jinwei Gu

Sifei Liu

Xueting Li

Sujin Jung

Pavlo Molchanov

Sunguk Jang

Youngdae Cho

Alejandro Jose Troccoli

Kari Pulli

Varun Jampani

Shalini Gupta

Junyoung Kim

Younjung Kim

Minyoung Lee

Deqing Sun

Changjung Kim

Yunson Yoo

Junghan Lee

Sanghee Park

Changhee Kim

Zhaoyang Lv

Jinhak Lee

Ilkwang Choi

Samarth Manoj Brahmbhatt

Benjamin David Eckart

Bosoo Kang

Suhyueon Park

Jaehong Choi

Michael S Brown

Ming-Yu Liu

Youngsoo Park

Changbum Lee

Xiaodong Yang

Chao Liu

Orazio Gallo

Junbeom Park

Jihwan Choe

Stephen Walter Tyree

Seungeon Ahn

Abdelrahman Abdelhamed

Junghyun Lee

Seung-Hwan Baek

Guilin Liu

Jonghwa Yim

Soumyadip Sengupta

Myungjae Lee

Jae Shin Yoon

Mark Boss

Chen Liu

Apparatuses, systems, and techniques are presented to determine visual properties of an object. In at least one embodiment, shape and visual properties of an object can be determined using differently illuminated images fed to a pipeline of neural networks.

Visual property determinations using one or more neural networks

Apparatuses, systems, and techniques to identify a shape or camera pose of a three-dimensional object from a two-dimensional image of the object. In at least one embodiment, objects are identified in an image using one or more neural networks that have been trained on objects of a similar category and a three-dimensional mesh template.

Self-supervised single-view 3D reconstruction via semantic consistency

A three-dimensional (3D) object reconstruction neural network system learns to predict a 3D shape representation of an object from a video that includes the object. The 3D reconstruction technique may be used for content creation, such as generation of 3D characters for games, movies, and 3D printing. When 3D characters are generated from video, the content may also include motion of the character, as predicted based on the video. The 3D object construction technique exploits temporal consistency to reconstruct a dynamic 3D representation of the object from an unlabeled video. Specifically, an object in a video has a consistent shape and consistent texture across multiple frames. Texture, base shape, and part correspondence invariance constraints may be applied to fine-tune the neural network system. The reconstruction technique generalizes well—particularly for non-rigid objects.

Three-dimensional object reconstruction from a video

An electronic apparatus according to various embodiments disclosed in the disclosure may comprise: a first housing rotatable about a first axis; a projector module coupled to a portion of the first housing and rotatable about a second axis perpendicular to the first axis; a vision sensor module rotatable about the first axis; a second housing; a first driving device disposed inside the first housing and capable of transmitting power to each of the projector module and the first housing; and a second driving device disposed inside the second housing and capable of transmitting power to the vision sensor module. According to the various embodiments, the projector module can rotate independently of the first housing in the direction of the second axis, and the vision sensor module can rotate independently of the first housing in the direction of the first axis.

Electronic apparatus including projector

A three-dimensional (3D) object reconstruction neural network system learns to predict a 3D shape representation of an object from a video that includes the object. The 3D reconstruction technique may be used for content creation, such as generation of 3D characters for games, movies, and 3D printing. When 3D characters are generated from video, the content may also include motion of the character, as predicted based on the video. The 3D object construction technique exploits temporal consistency to reconstruct a dynamic 3D representation of the object from an unlabeled video. Specifically, an object in a video has a consistent shape and consistent texture across multiple frames. Texture, base shape, and part correspondence invariance constraints may be applied to fine-tune the neural network system. The reconstruction technique generalizes well&#x2014;particularly for non-rigid objects.

A semiconductor device including an active region protruding from an upper surface of a substrate and extending in a first horizontal direction, at least two gate electrodes extending in a second horizontal direction and crossing the active region, the second horizontal direction crossing the first horizontal direction, a source/drain region in the active region between the gate electrodes may be provided. The source/drain region includes a recess region, an outer doped layer on an inner wall of the recess region, an intermediate doped layer on the outer doped layer, and an inner doped layer on the intermediate doped layer and filling the recess region. One of the outer doped layer or the intermediate doped layer includes antimony, and the inner doped layer includes phosphorous.

Semiconductor devices including source/drain regions having antimony doped layers

A semiconductor device including: an active pattern on a substrate, the active pattern including a recess, the recess having a 'V' shape; a growth prevention pattern on the recess; gate structures on portions of the active pattern at opposite sides of the recess; channels spaced apart from each other in a vertical direction perpendicular to an upper surface of the substrate, each of the channels extending through one of the gate structures; and a source/drain layer on the growth prevention pattern, the source/drain layer contacting the channels.

Semiconductor devices

A method for generating an image that includes at least one of a vignette effect or a grain effect corresponding to an input image may include obtaining the input image including at least one of the vignette effect or the grain effect; identifying at least one of a vignette parameter or a grain parameter of the input image; obtaining at least one of a vignette filter based on the vignette parameter or a grain layer based on the grain parameter; and generating the image that includes at least one of the vignette effect or the grain effect by applying at least one of the vignette filter or the grain layer to the image.

System and method for obtaining and applying a vignette filter and grain layer

Apparatuses, systems, and techniques are presented to perform monocular view synthesis of a dynamic scene. Single and multi-view depth information can be determined for a collection of images of a dynamic scene, and a blender network can be used to combine image features for foreground, background, and missing image regions using fused depth maps inferred form the single and multi-view depth information.

View synthesis for dynamic scenes

A neural network model receives color data for a sequence of images corresponding to a dynamic scene in three-dimensional (3D) space. Motion of objects in the image sequence results from a combination of a dynamic camera orientation and motion or a change in the shape of an object in the 3D space. The neural network model generates two components that are used to produce a 3D motion field representing the dynamic (non-rigid) part of the scene. The two components are information identifying dynamic and static portions of each image and the camera orientation. The dynamic portions of each image contain motion in the 3D space that is independent of the camera orientation. In other words, the motion in the 3D space (estimated 3D scene flow data) is separated from the motion of the camera.

Learning rigidity of dynamic scenes for three-dimensional scene flow estimation

A semiconductor device, including a silicon on insulator (SOI) substrate is disclosed. The device may include gate structures formed on the SOI substrate and being spaced apart from each other in a horizontal direction, and a plurality of channels spaced apart from each other in a vertical direction. Each of the channels may extend through each of the gate structures in the horizontal direction. The device may include a seed layer and a source/drain region. The source/drain region may be connected to the channels, and each sidewall of the source/drain region in the horizontal direction may have a concave-convex shape. The device may include a protruding portion of the source/drain region formed between the gate structures that protrudes in the horizontal direction compared to a non-protruding portion of the source/drain region formed between the channels.

An electronic device includes a sensor, a display, a processor operatively coupled to the sensor and the display, and a memory operatively coupled to the processor. The memory stores instructions that, when executed by the processor, causes the processor to identify a first body part of a user based on the at least one sensor, identify a virtual region spaced apart by a designated distance from the first body part, based on the identified first body part, identify a second body part distinct from the first body part within the virtual region, based on the at least one sensor, and change at least part of content displayed on the display based on a position of the second body part within the virtual region.

Electronic device and method for providing content based on the motion of the user

Inverse rendering estimates physical scene attributes (e.g., reflectance, geometry, and lighting) from image(s) and is used for gaming, virtual reality, augmented reality, and robotics. An inverse rendering network (IRN) receives a single input image of a 3D scene and generates the physical scene attributes for the image. The IRN is trained by using the estimated physical scene attributes generated by the IRN to reproduce the input image and updating parameters of the IRN to reduce differences between the reproduced input image and the input image. A direct renderer and a residual appearance renderer (RAR) reproduce the input image. The RAR predicts a residual image representing complex appearance effects of the real (not synthetic) image based on features extracted from the image and the reflectance and geometry properties. The residual image represents near-field illumination, cast shadows, inter-reflections, and realistic shading that are not provided by the direct renderer.

Inverse rendering of a scene from a single image

Growing community of inventors

Campbell, CA, United States of America

Kihwan Kim