A residual network is used to predict a set of residual motion vectors that provide additional motion data for portions of the frame for which motion vectors are not provided, such as animated textures, mirrored/reflected objects, and/or moving objects without motion information.

A graphics processor is provided that includes circuitry configured to generate auxiliary motion vectors for higher-order light-based effects such as shadows, objects reflecting in mirrors, waves in water or other liquids, glossy surfaces, or objects visible through transparent and/or refractive glass. The circuitry is configured to apply light path constraints to simplify the calculations used to generate the auxiliary motion vectors.

An apparatus to facilitate combined denoising and upscaling network with importance sampling in a graphics environment is disclosed. The apparatus includes set of processing resources including circuitry configured to: receive, at an input of a density map neural network, a sampled signal of a current frame and a reconstructed sample of the current frame; output, from the density map neural network, a prediction of a density map of samples based on the input of the current frame; provide the density map of samples to a sampler; reproject the density map of samples to a next frame; and apply the reprojected density map of samples to the next frame to generate a next sampled signal.

A method and system is provided to create animated faces using texture manipulation. A face template is provided to enable a user to define features of the face. A composite face is created from multiple layers that include a bottom layer, an animation layer, and a static layer. The composite face is animated by selectively animating one or more of the layers.

In one embodiment, a method includes adjusting parameters of a three-dimensional geometry corresponding to a first person to make the three-dimensional geometry represent a desired pose for the first person, accessing a neural texture encoding an appearance of the first person, generating a first rendered neural texture based on a mapping between (1) a portion of the three-dimensional geometry that is visible from a viewing direction and (2) the neural texture, generating a second rendered neural texture by processing the first rendered neural texture using a first neural network, determining normal information associated with the portion of the three-dimensional geometry that is visible from the viewing direction, and generating a rendered image for the first person in the desired pose by processing the second rendered neural texture and the normal information using a second neural network.

Systems, apparatuses, and methods for implementing gradient adaptive ringing control for image resampling are disclosed. A blending alpha calculation circuit generates a blending alpha value for a set of input pixels based on a normalized gradient calculated for the set of input pixels. The normalized gradient is a low-pass filtered gradient of the set of input pixels divided by a maximum gradient for the set of input pixels. The normalized gradient is passed through a mapping function so as to generate the blending alpha value. The mapping function is pre-tuned based on filter coefficients, video content type, pixel format, and so on. An interpolated pixel is generated for the set of input pixels by blending ringing free and ringing prone interpolation coefficients, or by blending results between ringing free and ringing prone interpolation filters, with the blending weight for each filter based on the blending alpha value.

In one embodiment, a user-navigable, three-dimensional (3-D) virtual room-based user interface for a home automation system is provided. Each user-navigable 3-D virtual room shows a substantially photo-realistic depiction of a corresponding physical room of the structure, including substantially photo-realistic depictions of boundaries of the physical room, furnishings present in the physical room, and devices present in the physical room that are under the control of the home automation system. A user may use explicit navigation commands or implicit actions to navigate within the user-navigable 3-D virtual room, moving a virtual camera in 3-D space to view the virtual room from different perspectives. By interacting with (e.g., touching, clicking on, etc.) substantially photorealistic depictions of the devices within the user-navigable 3-D virtual room, a user may is indicate changes to the state of corresponding devices in the physical room.

A method for generating an augmented reality image from first and second images, wherein at least a portion of at least one of the first and the second image is captured from a real scene, the method comprising: identifying a confidence region in which a confident determination as to which of the first and second image to render in that region of the augmented reality image can be made; identifying an uncertainty region in which it is uncertain as to which of the first and second image to render in that region of the augmented reality image; determining at least one blending factor value in the uncertainty region based upon a similarity between a first colour value in the uncertainty region and a second colour value in the confidence region; and generating an augmented reality image by combining, in the uncertainty region, the first and second images using the at least one blending factor value.

Apparatuses, systems, and techniques are presented to reconstruct one or more images. In at least one embodiment, one or more circuits are to use one or more neural networks to adjust one or more pixel blending weights.

Techniques for controlling patch-usage in image synthesis are described. In implementations, a curve is fitted to a set of sorted matching errors that correspond to potential source-to-target patch assignments between a source image and a target image. Then, an error budget is determined using the curve. In an example, the error budget is usable to identify feasible patch assignments from the potential source-to-target patch assignments. Using the error budget along with uniform patch-usage enforcement, source patches from the source image are assigned to target patches in the target image. Then, at least one of the assigned source patches is assigned to an additional target patch based on the error budget. Subsequently, an image is synthesized based on the source patches assigned to the target patches.