Tetavi Ltd.

Vsevolod Kagarlitsky

Michael Tamir

Michael Birnboim

Gilad Talmon

Shirley Keinan

Michael Priven

Yair Baruch

Michal Heker

David Dreizner

David Drezner

Amir Green

Yigal Eilam

Ofer Rubinstein

Gil Wohlstadter

Alik Mokeichev

A computer-implemented method of generating of a database for training a neural network configured for converting 2d images into 3d models comprising steps of: (a) obtaining 3d models; (b) rendering said 3d models in a 2d format from at least one view point; and (c) collecting pairs further comprising said rendered 2d image frame and said corresponding sampled 3d models each.

Method of training a neural network configured for converting 2D images into 3D models

The present invention generally pertains to systems, methods, and non-transitory processor-readable mediums for ensuring a match between geometry and texture when playing volumetric videos in a web browser.

Volumetric video in web browser

System and method for texturing a 3D surface using 2D images sourced from a plurality of imaging devices. System and method for applying a realistic texture to a model, based on texture found in one or more two-dimensional (2D) images of the object, with the texture covering the entire 3D model even if there are portions of the object that were invisible in the 2D image. System and method which does not require machine learning, is not incapable of blending between images, and which is not incapable of filling in portions of a 3D model that are invisible in the 2D image.

Surface texturing from multiple cameras

Methods and systems for generating free viewpoint videos (FVVs) based on images captured in a sports arena. A method includes projecting, onto objects within a filming area within the sports arena, a predefined pattern including a large set of features; generating, based on signals captured by each of a plurality of depth cameras, a point cloud for each depth camera, wherein the plurality of depth cameras is deployed in proximity to the filming area, wherein the captured signals are reflected off of the objects within the filming area; creating, based on the plurality of point clouds, a unified point cloud; meshing points in the unified point cloud to generate a three-dimensional (3D) model of the objects; texturing the 3D model based on images captured by the plurality of depth cameras; and rendering the textured 3D model as a FVV including a series of video frames with respect to a viewpoint.

System and method for rendering free viewpoint video for sport applications

Systems and methods for foreground/background separation and for studio production of a FVV. A method includes projecting, onto objects in a filming area within a studio, a predefined pattern including a large set of features; generating, based on signals reflected off of the objects and captured by a plurality of depth cameras deployed in proximity to the filming area, a local point cloud for each depth camera; separating, based on the local point clouds, between a background and a foreground of the filming area; creating, based on the local point clouds, a unified point cloud; meshing points in the unified point cloud to generate a 3D model of the objects; texturing the 3D model based on the separation and images captured by the depth cameras; and rendering the textured 3D model as a FVV including a series of video frames with respect to at least one viewpoint.

System and method for rendering free viewpoint video for studio applications

A method and system for improving a three-dimensional (3D) representation of objects using semantic data. The method comprises receiving an input data generated in response to captured video in a filming area; setting at least one parameter for each region in the input data; and generating a 3D representation based in part on the at least one parameter and semantic data associated with the input data.

Growing community of inventors

Ramat Gan, Israel

Vsevolod Kagarlitsky