Abstract: Machine-learning model retargeting techniques are described. In one example, training data is generated by extrapolating feedback data collected from entities. These techniques supports an ability to identify a wider range of thresholds and corresponding entities than those available in the feedback data. This also provides an opportunity to explore additional thresholds than those used in the past through extrapolating operations outside of a range used to define a segment, for which, the feedback data is captured. These techniques also support retargeting of a machine-learning model for a secondary label that is different than a primary label used to initially train the machine-learning model.

Abstract: In implementations of systems for image inversion using multiple latent spaces, a computing device implements an inversion system to generate a segment map that segments an input digital image into a first image region and a second image region and assigns the first image region to a first latent space and the second image region to a second latent space that corresponds to a layer of a convolutional neural network. An inverted latent representation of the input digital image is computed using a binary mask for the second image region. The inversion system modifies the inverted latent representation of the input digital image using an edit direction vector that corresponds to a visual feature. An output digital image is generated that depicts a reconstruction of the input digital image having the visual feature based on the modified inverted latent representation of the input digital image.

Abstract: Techniques are described that support automated generation of a digital document from digital videos using machine learning. The digital document includes textual components that describe a sequence of entity and action descriptions from the digital video. These techniques are usable to generate a single digital document based on a plurality of digital videos as well as incorporate user-specified constraints in the generation of the digital document.

Abstract: Embodiments are disclosed for eliminating typographical errors from an electronic document. The method may include obtaining an electronic document comprising a plurality of text paragraphs. The method may further include detecting a plurality of typographical errors in the plurality of text paragraphs. The method may further include indexing a set of error paragraphs, wherein each paragraph in the set of error paragraphs includes at least one typographical error. The method may further include determining a priority for each typographical error based on a magnitude of the typographical error. The method may further include adjusting one or more attributes of each paragraph in the set of error paragraphs based on the priority for each typographical error.

Abstract: Vector object path segment editing techniques are described that retain edibility of a path while supporting editing of a segment included within the path, individually and separately, without editing other segments of path. A vector object editing module first retrieves information on segments included in a path of a vector object. The vector object editing module then renders a selected segment separately from an adjacent segment based on the selected segment model. An editing operation is then applied to the selected segment as specified via the user interface, e.g., to change color, width, or other display characteristic. The vector object editing module then generates a joint between the edited segment and the adjacent segment to provide a transition between the segments that mimics inclusion as a single path that contains those segments.

Abstract: Techniques are described for error log anomaly detection. In an implementation, error logs from an application are processed to generate training data. The error logs, for instance, are processed to remove personal information and other data such as numerical strings. The processed error logs are converted into embeddings to generate the training data. The training data is utilized to train an anomaly detection model. For instance, as part of training the anomaly detection model, an anomaly threshold is defined based on a loss value determined from output of the anomaly detection model. Further error logs from the application are then processed by the trained anomaly detection model to determine which of the further error logs are error anomalies, such as based on comparing loss values for the further error logs to the anomaly threshold.

Abstract: Techniques and systems are provided for glyph-aware text selection. For instance, a glyph selection system can detect that a user has selected a glyph within a user interface. The glyph selection system can highlight the glyph and/or a region encompassing the glyph to communicate, to the user, that the glyph is selected. This highlighted region can be determined based on the shape and/or outline of the glyph. For example, the glyph selection system can determine bounds (e.g. coordinates) of the glyph in order to highlight a region within the user interface that fully encompasses the glyph and does not include portions of unselected glyphs. In some cases, the highlighted region may be rectangular. Alternatively, the highlighted region may be non-rectangular, such as a border defined by the outline of the glyph.

Abstract: Multi-modal machine-learning model training techniques for search are described that overcome conventional challenges and inefficiencies to support real time output, which is not possible in conventional training techniques. In one example, a search system is configured to support multi-modal machine-learning model training. This includes use of a preview mode and an expanded mode. In the preview mode, a preview segment is generated as part of real time training of a machine learning model. In the expanded mode, the preview segment is persisted as an expanded segment that is used to train and utilize an expanded machine-learning model as part of search.

Abstract: In implementations of systems for generating images for virtual try-on and pose transfer, a computing device implements a generator system to receive input data describing a first digital image that depicts a person in a pose and a second digital image that depicts a garment. Candidate appearance flow maps are computed that warp the garment based on the pose at different pixel-block sizes using a first machine learning model. The generator system generates a warped garment image by combining the candidate appearance flow maps as an aggregate per-pixel displacement map using a convolutional gated recurrent network. A conditional segment mask is predicted that segments portions of a geometry of the person using a second machine learning model. The generator system outputs a digital image that depicts the person in the pose wearing the garment based on the warped garment image and the conditional segmentation mask using a third machine learning model.

Abstract: Techniques for responsive video canvas generation are described to impart three-dimensional effects based on scene geometry to two-dimensional digital objects in a two-dimensional design environment. A responsive video canvas, for instance, is generated from input data including a digital video and scene data. The scene data describes a three-dimensional representation of an environment and includes a plurality of planes. A visual transform is generated and associated with each plane to enable digital objects to interact with the underlying scene geometry. In the responsive video canvas, an edit positioning a two-dimensional digital object with respect to a particular plane of the responsive video canvas is received. A visual transform associated with the particular plane is applied to the digital object and is operable to align the digital object to the depth and orientation of the particular plane. Accordingly, the digital object includes visual features based on the three-dimensional representation.

Abstract: In implementations of repeat object blending, a computing device implements a repeat object blending system, which is implemented to receive a digital image depicting a first object and a second object, where the first object is depicted as multiple instances of a repeated base object, and the second object is depicted as multiple instances of a visually different repeated base object. The repeat object blending system can identify visual characteristics of the first object and the second object. The repeat object blending system can then generate an intermediate object by blending one or more of the visual characteristics of the first object and one or more of the visual characteristics of the second object. The resulting intermediate object is a visual representation of the repeated base object blended with the visually different repeated base object.

Abstract: In implementations of text wrap modification using variable inset, a display screen of a device displays lines of text wrapped to an inset space maintained between an object boundary and the lines of text. The device implements a text wrap modification module to determine that a penalty value associated with a line of text is reduced if the line of text is extended to include one or more words from a subsequent line of text, determine that the one or more words fit within an additional space for the line of text based on a variable overlap of the line of text into the inset space, and display the one or more words from the line of text as extended to include the one or more words from the subsequent line of text.

Abstract: In implementations of systems for cloud-based resource allocation using meters, a computing device implements a resource system to receive resource data describing an amount of cloud-based resources reserved for consumption by client devices during a period of time and a total amount of cloud-based resources consumed by the client devices during the period of time. The resource system determines a consumption distribution using each meter included in a set of meters. Each of the consumption distributions allocates a portion of the total amount of the cloud-based resources consumed to each client device of the client devices. A particular meter used to determine a particular consumption distribution is selected based on a Kendall Tau coefficient of the particular consumption distribution. An amount of cloud-based resources to allocate for a future period of time is estimated using the particular meter and an approximate Shapley value.

Abstract: Keyword localization digital image search techniques are described. These techniques support an ability to indicate “where” a corresponding keyword is to be expressed with respect to a layout in a respective digital image resulting from a search query. The search query may also include an indication of a size of the keyword as expressed in the digital image, a number of instances of the keyword, and so forth. Additionally, the techniques and systems as described herein support real time search through use of keyword signatures.

Abstract: Techniques for content-aware font recommendations include obtaining an electronic document comprising an image and text. The image is processed using one or more convolutional neural networks to determine one or more image tags. The image tags are mapped to one or more font tags using a user map, a designer map, or one or more contextual synonyms of the image tags. A font to recommend for the electronic document is then determined using the one or more font tags.

Abstract: In implementations of systems for generating and applying editing presets, a computing device implements a preset system to detect objects depicted in a digital image that is displayed in a user interface of an application for editing digital content. Input data is received describing an edited region of the digital image and properties of an editing operation performed in the edited region. The preset system identifies a particular detected object of the detected objects based on a bounding box of the particular detected object and an area of the edited region. An additional digital image is edited by applying the properties of the editing operation to a detected object that is depicted in the additional digital image based on a classification of the detected object and a classification of the particular detected object.

Abstract: Techniques and systems are described for generating location based photo discovery suggestions. Generally, photo discovery data is generated and utilized to form discovery suggestions that identify suggested locations for capturing photographs, as well as other capture-related information that is presented to assist a user in capturing photographs of interest to the user. A discovery suggestion, for example, not only identifies a location of potential photographic interest to a user, but also includes information such as camera settings and suggested camera equipment for capturing photos at the location. A discovery suggestion thus guides a user as to how to maximize a likelihood that a digital image captured by the user includes subject matter of interest to the user and is also visually pleasing.

Abstract: Vector object transformation techniques are described that support generation of a transformed vector object based on a first vector object and a second vector object. A plurality of paths for a first and second vector object, for instance, are generated. Corresponding paths are determined by detecting which of the plurality of paths from the first vector object correspond to which of the plurality of paths from the second vector object. A mapping of control points between the first and second vector objects is generated. Using the mapping, a transformation of the first vector object is generated by adjusting one or more control points of the first vector object. As a result, the transformed vector object includes visual characteristics based on both the first vector object and the second vector object.

Abstract: Directional propagation editing techniques are described, in one example, a digital image, a depth map, and a direction are obtained by an image editing system. The image editing system then generates features. To do so, the image editing system generates features from the digital image and the depth map for each pixel based on the direction, e.g., until an edge of the digital image is reached. In an implementation, instead of storing a value of the depth directly, a ratio is stored based on a depth in the depth map and a depth of a point along the direction. The image editing system then forms a feature volume using the features, e.g., as three dimensionally stacked features. The feature volume is employed by the image editing system as part of editing the digital image to form an edited digital image.

Abstract: Embodiments are disclosed for performing content linting in a graphic design system. A method of content linting includes receiving a selection of a content type to be generated, receiving a selection of a location in a digital canvas to place content of the content type, determining a placement context associated with the location in the digital canvas, identifying one or more content rules to the content based on a static analysis of the placement context, and generating, using one or more machine learning models, content of the selected content type at the location in the digital canvas using the one or more content rules.