Abstract: Certain embodiments involve flow-based color transfers from a source graphic to target graphic. For instance, a palette flow is computed that maps colors of a target color palette to colors of the source color palette (e.g., by minimizing an earth-mover distance with respect to the source and target color palettes). In some embodiments, such color palettes are extracted from vector graphics using path and shape data. To modify the target graphic, the target color from the target graphic is mapped, via the palette flow, to a modified target color using color information of the source color palette. A modification to the target graphic is performed (e.g., responsive to a preview function or recoloring command) by recoloring an object in the target color with the modified target color.

Abstract: Techniques for intuitive modifications of digital graphics in a digital media environment are described. For example, a digital graphics creation system accesses vector artwork including a vector object, such as a Bezier curve. The digital graphics creation system receives user inputs, including a user input defining handles on the vector object and a user input interacting with the handles indicating a desired change to the vector object. The digital graphics creation system modifies the vector artwork, including the vector object, by accounting for topology of the vector object and maintaining connections between connected segments of the vector object. The digital graphics creation system outputs the modified vector artwork, including the vector object, such as in a user interface.

Abstract: In implementations of weighted color palette generation, one or more computing devices implement a generation system which receives input data including an input color palette. A first machine learning model receives the input color palette and generates an unweighted color palette based on the input color palette. A second machine learning model receives the generated unweighted color palette and generates a weighted color palette based on the generated unweighted color palette. The generation system renders the weighted color palette in a user interface.

Abstract: In implementations of segmenting objects in vector graphics images, an object segmentation system can obtain points that identify an object in a vector graphics image, and determine a region of interest in the image that includes the object based on the points that identify the object. The object segmentation system can generate a heat map from the points that identify the object in the image, and a rasterized region from rasterizing the region of interest. The object segmentation system can generate a mask from the rasterized region and the heat map, the mask identifying pixels of the object in the rasterized region, and determine, from the mask, paths of the vector graphics corresponding to the object.

Abstract: Generation of a multi-state symbol from an input graphic object is described. A multi-state graphic symbol system generates an outline and a base mesh for a graphic object. The multi-state graphic symbol system then defines graphic manipulation handles relative to the base mesh and deforms the base mesh by altering a state of the handles. Vectors describing initial positions and final positions of the handles are generated and stored with the outline and base mesh to define the graphic object's multi-state symbol. Additional poses can be generated by adding and/or modifying other handles, and each additional pose is stored as a vector in the multi-state symbol. Additional poses of the graphic object can be generated by interpolating between different vectors of the multi-state symbol. The multi-state graphic symbol system additionally enables for an interpolated pose to be generated based on separate user-defined paths for different handles of the multi-state symbol.

Abstract: Techniques are disclosed for generating a vector image from a raster image, where the raster image is, for instance, a photographed or scanned version of a hand-drawn sketch. While drawing a sketch, an artist may perform multiple strokes to draw a line, and the resultant raster image may have adjacent or partially overlapping salient and non-salient lines, where the salient lines are representative of the artist's intent, and the non-salient (or auxiliary) lines are formed due to the redundant strokes or otherwise as artefacts of the creation process. The raster image may also include other auxiliary features, such as blemishes, non-white background (e.g., reflecting the canvas on which the hand-sketch was made), and/or uneven lighting. In an example, the vector image is generated to include the salient lines, but not the non-salient lines or other auxiliary features. Thus, the generated vector image is a cleaner version of the raster image.

Abstract: Certain embodiments involve flow-based color transfers from a source graphic to target graphic. For instance, a palette flow is computed that maps colors of a target color palette to colors of the source color palette (e.g., by minimizing an earth-mover distance with respect to the source and target color palettes). In some embodiments, such color palettes are extracted from vector graphics using path and shape data. To modify the target graphic, the target color from the target graphic is mapped, via the palette flow, to a modified target color using color information of the source color palette. A modification to the target graphic is performed (e.g., responsive to a preview function or recoloring command) by recoloring an object in the target color with the modified target color.

Abstract: A vector representation of a sketch is generated from a raster representation of the sketch. The vector representation is an abstraction of the sketch that captures the intent or desire of the user rather than a faithful reconstruction of what the user actually drew. A sketch includes one or more components. For each component, pixels of the component are clustered into multiple proxies. The multiple proxies are merged together based on their similarities and proximity to each other, resulting in a set of merged proxies. For each component, pixels within each of the merged proxies as well as merged proxies within the component are sorted into a natural drawing order. A set of geometric constructs representing the sketch are then generated by fitting one or more geometric constructs (e.g., circular arcs and line segments) onto the merged proxies.

Abstract: Systems and techniques are described herein for generating a triangle mesh for an image represented by curves (e.g., Bezier segments). An outline of an image is determined and reduced to a set of connected polylines that are efficiently represented in an edge list. A triangle mesh is generated based on the edge list, rather than by directly sampling the curves of the image and using the samples as vertices of triangles. Thus, the triangle mesh is generated with a number of triangles independent from a number of curves representing the image. Samples of the curves are bound to the triangle mesh by representing the samples with barycentric coordinates with respect to a triangle in the mesh. Hence, once a mesh is deformed, locations of the samples are determined from the barycentric coordinates and triangles in the deformed mesh, and used to reconstruct the curves of the deformed image.

Abstract: Color handles are generated based on digital image color gradients using machine learning. In order to train a neural network of a generator module of a GAN system, a training digital image is used by the neural network of the generator module to generate a candidate color handle. The discriminator module of the GAN system then compares the candidate color handle with a ground truth color handle of the training digital image using a loss function as part of machine learning, and uses a result of this comparison to train the neural network of the generator module.

Abstract: Automatic bone handle generation for efficient manipulation of a graphic object is described. Initially, a request is received to generate bone handles for a graphic object. Responsive to the request, a bone handle generation system determines an outline of the graphic object. Based on the outline, this system then generates a mesh corresponding to the graphic object. The bone handle generation system contracts this mesh inward to form a contracted mesh. By then collapsing edges of the contracted mesh, the bone handle generation system generates a set of connected line segments forming a skeleton of the graphic object. The bone handle generation system reduces the set of connected line segments by removing line segments and by merging vertices that connect different line segments. Based on this reduction, the bone handle generation system obtains the bone handles, which are controllable with user input to deform corresponding portions of the graphic object.

Abstract: Generation of a multi-state symbol from an input graphic object is described. A multi-state graphic symbol system generates an outline and a base mesh for a graphic object. The multi-state graphic symbol system then defines graphic manipulation handles relative to the base mesh and deforms the base mesh by altering a state of the handles. Vectors describing initial positions and final positions of the handles are generated and stored with the outline and base mesh to define the graphic object's multi-state symbol. Additional poses can be generated by adding and/or modifying other handles, and each additional pose is stored as a vector in the multi-state symbol. Additional poses of the graphic object can be generated by interpolating between different vectors of the multi-state symbol. The multi-state graphic symbol system additionally enables for an interpolated pose to be generated based on separate user-defined paths for different handles of the multi-state symbol.

Abstract: Systems and techniques are described for colorizing vector images. Color raster images that correspond to the vector images are used to perform a color mapping to colorize the vector images. The vector images are represented using non-overlapping planar arrangements of faces of the vector images, so that the color mapping may be performed with respect to the faces. The faces may be processed in parallel, to further facilitate a speed and scalability of the describes processes and results.

Abstract: A selection of a key path of a vector image displayed using a graphical user interface (GUI) may be received, via the GUI. At least one candidate path of the vector image is identified. A pairwise comparison of the key path with the at least one candidate path is executed, the pairwise comparison including characterization of a translation, scaling, and rotation of the at least one candidate path with respect to the key path. Based on the pairwise comparison, it is determined that the at least one candidate path is within a similarity threshold defined with respect to the key path. A visual indicator of the at least one candidate path within the GUI, identifying the at least one candidate path as being within the similarity threshold, may be provided.

Abstract: Automatic bone handle generation for efficient manipulation of a graphic object is described. Initially, a request is received to generate bone handles for a graphic object. Responsive to the request, a bone handle generation system determines an outline of the graphic object. Based on the outline, this system then generates a mesh corresponding to the graphic object. The bone handle generation system contracts this mesh inward to form a contracted mesh. By then collapsing edges of the contracted mesh, the bone handle generation system generates a set of connected line segments forming a skeleton of the graphic object. The bone handle generation system reduces the set of connected line segments by removing line segments and by merging vertices that connect different line segments. Based on this reduction, the bone handle generation system obtains the bone handles, which are controllable with user input to deform corresponding portions of the graphic object.

Abstract: Systems and techniques are described for determining image handle locations. An image is provided to a neural network as input, and the neural network translates the input image to an output image that includes clusters of pixels against a background that have intensities greater than an intensity of the background and that indicate candidate handle locations. Intensities of clusters of pixels in an output image are compared to a threshold intensity level to determine a set of the clusters of pixels satisfying an intensity constraint. The threshold intensity level can be user-selectable, so that a user can control a density of handles. A handle location for each cluster of the set of clusters is determined from a centroid of each cluster. Handle locations include a coordinate for the handle location and an attribute classifying a degree of freedom for a handle at the handle location.

Abstract: Color handles are generated based on digital image color gradients using machine learning. In order to train a neural network of a generator module of a GAN system, a training digital image is used by the neural network of the generator module to generate a candidate color handle. The discriminator module of the GAN system then compares the candidate color handle with a ground truth color handle of the training digital image using a loss function as part of machine learning, and uses a result of this comparison to train the neural network of the generator module.

Abstract: Systems and techniques are described for determining image handle locations. An image is provided to a neural network as input, and the neural network translates the input image to an output image that includes clusters of pixels against a background that have intensities greater than an intensity of the background and that indicate candidate handle locations. Intensities of clusters of pixels in an output image are compared to a threshold intensity level to determine a set of the clusters of pixels satisfying an intensity constraint. The threshold intensity level can be user-selectable, so that a user can control a density of handles. A handle location for each cluster of the set of clusters is determined from a centroid of each cluster. Handle locations include a coordinate for the handle location and an attribute classifying a degree of freedom for a handle at the handle location.

Abstract: A selection of a key path of a vector image displayed using a graphical user interface (GUI) may be received, via the GUI. At least one candidate path of the vector image is identified. A pairwise comparison of the key path with the at least one candidate path is executed, the pairwise comparison including characterization of a translation, scaling, and rotation of the at least one candidate path with respect to the key path. Based on the pairwise comparison, it is determined that the at least one candidate path is within a similarity threshold defined with respect to the key path. A visual indicator of the at least one candidate path within the GUI, identifying the at least one candidate path as being within the similarity threshold, may be provided.

Abstract: There is disclosed a system and method for colorizing vector graphic objects in a digital medium environment. The system comprises a processing unit and a deep neural network of the processing unit, in which the deep neural network includes a generator. The processing unit receives a non-colorized vector image and converts the non-colorized vector image to a non-colorized raster image. The deep neural network generates a colorized raster image from the non-colorized raster image. The generator processes the non-colorized raster image using an extended number of convolutional layers and residual blocks to add skip connections between at least two of the convolutional layers. The processing unit converts the colorized raster image to a colorized vector image.