Adobe, Inc.

Czech Technical University in Prague

Brno University of Technology

Michal Lukác

Elya Shechtman

Jan Brejcha

Nathan Aaron Carr

Kalyan Krishna Sunkavalli

Zhaowen Wang

Oliver Wang

Matthew David Fisher

Vineet Batra

Ankit Phogat

Zhili Chen

Kevin Wampler

Yannick Hold-Geoffroy

Niloy Jyoti Mitra

Wilmot Wei-Mau Li

Eric Joel Stollnitz

Daniel Max Kaufman

Daniel Sýkora

Michael Gharbi

Souymodip Chakraborty

Vidya Narayanan

Amanda Paige Ghassaei

Martin Cadik

Martin Cadík

Chi Cheng Hsu

David Futschik

Michaël Yanis Gharbi

Chinthala Pradyumna Yanis Reddy

Embodiments are disclosed for reconstructing linear gradients from an input image that can be applied to another image. In particular, in one or more embodiments, the disclosed systems and methods comprise receiving a raster image, the raster image including a representation of a linear color gradient. The disclosed systems and methods further comprise determining a vector representing a direction of the linear color gradient. The disclosed systems and methods further comprise analyzing pixel points along the direction of the linear color gradient to compute color stops of the linear color gradient. The disclosed systems and methods further comprise generating an output color gradient vector with the computed color stops of the linear color gradient, the output color gradient vector to be applied to a vector graphic.

Reconstructing linear gradients

Generating vector representations of visual objects is leveraged in a digital medium environment. For instance, a raster-based visual input object is encoded into a global latent code and individual path latent codes for visual components of the raster visual object are extracted from the global latent code. The path latent codes are decoded and used to generate vector representations of the original raster versions of the visual components. The vector representations are rasterized and composited to generate an output object that simulates a visual appearance of the input object.

Generating vector representations of visual objects

Embodiments are disclosed for translating an image from a source visual domain to a target visual domain. In particular, in one or more embodiments, the disclosed systems and methods comprise a training process that includes receiving a training input including a pair of keyframes and an unpaired image. The pair of keyframes represent a visual translation from a first version of an image in a source visual domain to a second version of the image in a target visual domain. The one or more embodiments further include sending the pair of keyframes and the unpaired image to an image translation network to generate a first training image and a second training image. The one or more embodiments further include training the image translation network to translate images from the source visual domain to the target visual domain based on a calculated loss using the first and second training images.

Neural network for image style translation

Embodiments are disclosed for receiving a target shape. The method may further include initializing a gradient mesh to a vector graphic having at least one node. The method may further include performing a constrained optimization of the vector graphic based on the target shape. The method may further include generating a stress metric based on a comparison of the constrained optimization and the target shape. The method may further include determining one or more unconstrained candidate vector graphics based on the stress metric. The method may further include selecting an improved vector graphic from the one or more unconstrained candidate vector graphics. The method may further include mapping the vector graphic to the improved vector graphic. The method may further include optimizing the improved vector graphic based on the target shape.

Continuous optimization of discrete parameters using a unified stress indicator

Methods and systems are provided for facilitating large-scale augmented reality in relation to outdoor scenes using estimated camera pose information. In particular, camera pose information for an image can be estimated by matching the image to a rendered ground-truth terrain model with known camera pose information. To match images with such renders, data driven cross-domain feature embedding can be learned using a neural network. Cross-domain feature descriptors can be used for efficient and accurate feature matching between the image and the terrain model renders. This feature matching allows images to be localized in relation to the terrain model, which has known camera pose information. This known camera pose information can then be used to estimate camera pose information in relation to the image.

Large-scale outdoor augmented reality scenes using camera pose based on learned descriptors

Image modification using detected symmetry is described. In example implementations, an image modification module detects multiple local symmetries in an original image by discovering repeated correspondences that are each related by a transformation. The transformation can include a translation, a rotation, a reflection, a scaling, or a combination thereof. Each repeated correspondence includes three patches that are similar to one another and are respectively defined by three pixels of the original image. The image modification module generates a global symmetry of the original image by analyzing an applicability to the multiple local symmetries of multiple candidate homographies contributed by the multiple local symmetries. The image modification module associates individual pixels of the original image with a global symmetry indicator to produce a global symmetry association map. The image modification module produces a manipulated image by manipulating the original image under global symmetry constraints imposed by the global symmetry association map.

Image modification using detected symmetry

A neighboring panel view is leveraged in a digital medium environment. Initially, a view generation system receives a selection of panels of a three-dimensional (3D) assembled representation of an object, such as to apply digital graphics to the 3D assembled representation. The view generation system also receives a selection of an option to display a neighboring panel view, which is a two-dimensional (2D) flattened view local to the selection. In particular, the neighboring panel view displays an arrangement of the selected panels of the object and their neighboring panels. The view generation system arranges the selected panels in this arrangement as those panels are arranged in the 3D assembled representation. The view generation system also arranges the neighboring panels in the neighboring panel view adjacent to the selected panels with which they are adjacent on the object.

Neighboring panel view

A user selects a set of photographs from a trip through an environment that he or she desires to present to other people. A collection of photographs, including the set of photographs captured during the trip optionally augmented with additional photographs obtained from another collection, are combined with a terrain model (e.g., a digital elevation model) to extract information regarding the geographic location of each of the photographs within the environment. The collection of photographs are analyzed, considering their geolocation information as well as the photograph content to register the photographs relative to one another. This information for the photographs is compared to the terrain model in order to accurately position the viewpoint for each photograph within the environment. A presentation of the selected photographs within the environment is generated that displays both the selected photographs and synthetic data filled in beyond the edges of the selected photographs.

Generating immersive trip photograph visualizations

Growing community of inventors

San Jose, CA, United States of America

Michal Lukác