Abstract: In some examples, a computing system access a set of registered three-dimensional (3D) digital shapes. The set of registered 3D digital shapes are registered to a shape template. The computing system determines a linear model for an estimate of the shape space using a first subset of the set of registered 3D digital shapes. The computing system then determines a nonlinear deformation model for the shape space using a second subset of the set of registered 3D digital shapes. An unregistered shape can be registered to the shape space using the linear model and the nonlinear deformation model. The registration can be added to the set of registered 3D digital shapes to update the estimate of the shape space if a shape distance between the registration and the unregistered shape is below a threshold value.

Abstract: Systems and methods are described for rendering garments. The system includes a first machine learning model trained to generate coarse garment templates of a garment and a second machine learning model trained to render garment images. The first machine learning model generates a coarse garment template based on position data. The system produces a neural texture for the garment, the neural texture comprising a multi-dimensional feature map characterizing detail of the garment. The system provides the coarse garment template and the neural texture to the second machine learning model trained to render garment images. The second machine learning model generates a rendered garment image of the garment based on the coarse garment template of the garment and the neural texture.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and method that utilize a character animation neural network informed by motion and pose signatures to generate a digital video through person-specific appearance modeling and motion retargeting. In particular embodiments, the disclosed systems implement a character animation neural network that includes a pose embedding model to encode a pose signature into spatial pose features. The character animation neural network further includes a motion embedding model to encode a motion signature into motion features. In some embodiments, the disclosed systems utilize the motion features to refine per-frame pose features and improve temporal coherency. In certain implementations, the disclosed systems also utilize the motion features to demodulate neural network weights used to generate an image frame of a character in motion based on the refined pose features.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media for accurately and flexibly generating scalable fonts utilizing multi-implicit neural font representations. For instance, the disclosed systems combine deep learning with differentiable rasterization to generate a multi-implicit neural font representation of a glyph. For example, the disclosed systems utilize an implicit differentiable font neural network to determine a font style code for an input glyph as well as distance values for locations of the glyph to be rendered based on a glyph label and the font style code. Further, the disclosed systems rasterize the distance values utilizing a differentiable rasterization model and combines the rasterized distance values to generate a permutation-invariant version of the glyph corresponding glyph set.

Abstract: Certain aspects and features of this disclosure relate to neural network based 3D object surface mapping. In one example, a first representation of a first surface of a first 3D object and a second representation of a second surface of a second 3D object are produced. A surface mapping function is generated for mapping the first surface to the second surface. The surface mapping function is defined the representations and by a neural network model configured to map a first 2D representation of the first surface to a second 2D representation of the second surface. One or more features of the a first 3D mesh on the first surface can be applied to a second 3D mesh on the second surface using the surface mapping function to produce a modified second surface, which can be rendered through a user interface.

Abstract: Systems and methods are described for rendering garments. The system includes a first machine learning model trained to generate coarse garment templates of a garment and a second machine learning model trained to render garment images. The first machine learning model generates a coarse garment template based on position data. The system produces a neural texture for the garment, the neural texture comprising a multi-dimensional feature map characterizing detail of the garment. The system provides the coarse garment template and the neural texture to the second machine learning model trained to render garment images. The second machine learning model generates a rendered garment image of the garment based on the coarse garment template of the garment and the neural texture.

Abstract: Certain aspects and features of this disclosure relate to neural network based 3D object surface mapping. In one example, a first representation of a first surface of a first 3D object and a second representation of a second surface of a second 3D object are produced. A surface mapping function is generated for mapping the first surface to the second surface. The surface mapping function is defined the representations and by a neural network model configured to map a first 2D representation of the first surface to a second 2D representation of the second surface. One or more features of the a first 3D mesh on the first surface can be applied to a second 3D mesh on the second surface using the surface mapping function to produce a modified second surface, which can be rendered through a user interface.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and method that utilize a character animation neural network informed by motion and pose signatures to generate a digital video through person-specific appearance modeling and motion retargeting. In particular embodiments, the disclosed systems implement a character animation neural network that includes a pose embedding model to encode a pose signature into spatial pose features. The character animation neural network further includes a motion embedding model to encode a motion signature into motion features. In some embodiments, the disclosed systems utilize the motion features to refine per-frame pose features and improve temporal coherency. In certain implementations, the disclosed systems also utilize the motion features to demodulate neural network weights used to generate an image frame of a character in motion based on the refined pose features.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media for accurately and flexibly generating scalable fonts utilizing multi-implicit neural font representations. For instance, the disclosed systems combine deep learning with differentiable rasterization to generate a multi-implicit neural font representation of a glyph. For example, the disclosed systems utilize an implicit differentiable font neural network to determine a font style code for an input glyph as well as distance values for locations of the glyph to be rendered based on a glyph label and the font style code. Further, the disclosed systems rasterize the distance values utilizing a differentiable rasterization model and combines the rasterized distance values to generate a permutation-invariant version of the glyph corresponding glyph set.

Abstract: A computer-implemented method of creating a two dimensional code comprises providing a two dimensional code comprising a cell, providing a picture comprising a patch corresponding to the cell of the two dimensional code, providing a plurality of cell patterns different from each other, wherein each cell pattern comprises a plurality of sub-cells, and determining one of the plurality of cell patterns for the cell of the two dimensional code according to the patch of the picture.

Abstract: An image processing method includes a segmentation step that segments an input image into a plurality of regions by using an automatic segmentation algorithm, and a computation step that calculates a saliency value of one region of the plurality of segmented regions by using a weighted sum of color differences between the one region and all other regions. Accordingly, it is possible to automatically analyze visual saliency regions in an image, and a result of analysis can be used in application areas including significant object segmentation, object recognition, adaptive image compression, content-aware image resizing, and image retrieval.

Abstract: A computer-implemented method of creating a two dimensional code comprises providing a two dimensional code comprising a cell, providing a picture comprising a patch corresponding to the cell of the two dimensional code, providing a plurality of cell patterns different from each other, wherein each cell pattern comprises a plurality of sub-cells, and determining one of the plurality of cell patterns for the cell of the two dimensional code according to the patch of the picture.

Abstract: Systems and methods for enhancing symmetry in 2D and 3D objects are disclosed. At least some embodiments include a computer-readable storage medium including software (executable on a processor) to symmetrize a modeled physical object that causes the processor to identify a plurality of clusters (each including a plurality of symmetric point pairs each derived from a plurality of sampled surface points of the object), and to calculate and apply each of a first plurality of displacement value pairs to corresponding sample positions of the symmetric point pairs within at least one cluster, increasing the symmetry of the cluster. The software further causes the processor to calculate a second plurality of displacement value pairs, to contract the cluster using the second plurality of displacement value pairs, to merge two or more clusters within the transformation space, and to present a graphical representation of the symmetrized modeled physical object to a user.

Abstract: An image processing method includes a segmentation step that segments an input image into a plurality of regions by using an automatic segmentation algorithm, and a computation step that calculates a saliency value of one region of the plurality of segmented regions by using a weighted sum of color differences between the one region and all other regions. Accordingly, it is possible to automatically analyze visual saliency regions in an image, and a result of analysis can be used in application areas including significant object segmentation, object recognition, adaptive image compression, content-aware image resizing, and image retrieval.

Abstract: Systems and methods for enhancing symmetry in 2D and 3D objects are disclosed. At least some embodiments include a computer-readable storage medium including software (executable on a processor) to symmetrize a modeled physical object that causes the processor to identify a plurality of clusters (each including a plurality of symmetric point pairs each derived from a plurality of sampled surface points of the object), and to calculate and apply each of a first plurality of displacement value pairs to corresponding sample positions of the symmetric point pairs within at least one cluster, increasing the symmetry of the cluster. The software further causes the processor to calculate a second plurality of displacement value pairs, to contract the cluster using the second plurality of displacement value pairs, to merge two or more clusters within the transformation space, and to present a graphical representation of the symmetrized modeled physical object to a user.

Abstract: A computer system includes a memory and a processor. The memory stores a program to cause the processor to provide wavelet coefficients that indicate an image. The processor represent each wavelet coefficient as a collection of ordered bits, and the processor codes the bits of each order to indicate zerotree roots that are associated with the order.

Abstract: A computer system includes a memory and a processor. The memory stores a program to cause the processor to provide error data that indicate motion in an image. The processor represent each error signal as a collection of ordered bits, and the processor codes the bits of each order to indicate zerotree roots that are associated with the order.

Abstract: A computer system includes a memory and a processor. The memory stores a program to cause the processor to provide error data that indicate motion in an image. The processor represent each error signal as a collection of ordered bits, and the processor codes the bits of each order to indicate zerotree roots that are associated with the order.

Abstract: In accordance with one embodiment of the invention, a method of compressing a color image includes the following. The wavelet transform of each of the respective color planes of the color image is computed. One of the respective wavelet transformed color plane frames is encoded. For the other two respective wavelet transformed color plane frames, a prediction coefficient from at least one of subbands of each color plane thereof is computed. In accordance with another embodiment of the invention, a method of decompressing a compressed color image includes the following. The compressed color image includes at least an encoded frame for one color plane of the color image and prediction coefficients for the other two color plane frames of the color image. The one color plane frame of the color image is reconstructed from the encoded frame. The other two color plane frames of the color image are at least partially reconstructed from the reconstructed one color plane frame and the prediction coefficients.

Abstract: A computer system includes a memory and a processor. The memory stores a program to cause the processor to provide wavelet coefficients that indicate an image. The processor represent each wavelet coefficient as a collection of ordered bits, and the processor codes the bits of each order to indicate zerotree roots that are associated with the order.