Abstract: One embodiment provides a graphics processor comprising a block of execution resources, a cache memory, a cache memory prefetcher, and circuitry including a programmable neural network unit, the programmable neural network unit comprising a network hardware block including circuitry to perform neural network operations and activation operations for a layer of a neural network, the programmable neural network unit addressable by cores within the block of graphics cores and the neural network hardware block configured to perform operations associated with a neural network configured to determine a prefetch pattern for the cache memory prefetcher.

Abstract: One embodiment provides for a graphics processor comprising a block of graphics compute units, a graphics processor pipeline coupled to the block of graphics compute units, and a programmable neural network unit including one or more neural network hardware blocks. The programmable neural network unit is coupled with the block of graphics compute units and the graphics processor pipeline. The one or more neural network hardware blocks include hardware to perform neural network operations and activation operations for a layer of a neural network. The programmable neural network unit can configure settings of one or more hardware blocks within the graphics processor pipeline based on a machine learning model trained to optimize performance of a set of workloads.

Abstract: Embodiments are directed to neural network processing for multi-object three-dimensional (3D) modeling. An embodiment of a computer-readable storage medium includes executable computer program instructions for obtaining data from multiple cameras, the data including multiple images, and generating a 3D model for 3D imaging based at least in part on the data from the cameras, wherein generating the 3D model includes one or more of performing processing with a first neural network to determine temporal direction based at least in part on motion of one or more objects identified in an image of the multiple images or performing processing with a second neural network to determine semantic content information for an image of the multiple images.

Abstract: An apparatus to facilitate deep learning based selection of samples for adaptive supersampling is disclosed. The apparatus includes one or more processing elements to: receive training data comprising input tiles and corresponding supersampling values for the input tiles, wherein each input tile comprises a plurality of pixels, and train, based on the training data, a machine learning model to identify a level of supersampling for a rendered tile of pixels.

Abstract: One embodiment provides for a graphics processor comprising a block of graphics compute units, a graphics processor pipeline coupled to the block of graphics compute units, and a programmable neural network unit including one or more neural network hardware blocks. The programmable neural network unit is coupled with the block of graphics compute units and the graphics processor pipeline. The one or more neural network hardware blocks include hardware to perform neural network operations and activation operations for a layer of a neural network. The programmable neural network unit can configure settings of one or more hardware blocks within the graphics processor pipeline based on a machine learning model trained to optimize performance of a set of workloads.

Abstract: An apparatus to facilitate deepfake detection models utilizing subject-specific libraries is disclosed. The apparatus includes one or more processors to store a plurality of deepfake detection models corresponding to a plurality of subjects of interest; receive a query to identify whether data pertaining to a target subject of interest is a deepfake, the target subject of interest comprised in the plurality of subjects of interest and associated with a subject identifier (ID); identify a deepfake detection model corresponding to the subject ID; extract features for deepfake detection from the data; input the extracted features to the identified deepfake detection model corresponding to the subject ID; and responsive to an output of the deepfake detection model exceeding a determined deepfake threshold, generate a notification, in response to the query, indicating a possible deepfake attack corresponding to the target subject of interest.

Abstract: An apparatus to facilitate deep learning based selection of samples for adaptive supersampling is disclosed. The apparatus includes one or more processing elements to: receive training data comprising input tiles and corresponding supersampling values for the input tiles, wherein each input tile comprises a plurality of pixels, and train, based on the training data, a machine learning model to identify a level of supersampling for a rendered tile of pixels.

Abstract: One embodiment provides for a graphics processor comprising a block of graphics compute units, a graphics processor pipeline coupled to the block of graphics compute units, and a programmable neural network unit including one or more neural network hardware blocks. The programmable neural network unit is coupled with the block of graphics compute units and the graphics processor pipeline. The one or more neural network hardware blocks include hardware to perform neural network operations and activation operations for a layer of a neural network. The programmable neural network unit can configure settings of one or more hardware blocks within the graphics processor pipeline based on a machine learning model trained to optimize performance of a set of workloads.

Abstract: Technologies for virtual attribute assignment include a compute device. The compute device is configured to receive an attribute assignment command from a user and analyze the attribute assignment command to determine a user-selected virtual object, a user-referenced attribute of the user-selected virtual object, a user-selected real object, and a user-referenced attribute of the user-selected real object. Based on the attribute assignment command, the compute device is further configured to determine a state of the user-referenced attribute of the user-selected real object and update a state of the user-referenced attribute of the user-selected virtual object based on the state of the user-referenced attribute of the user-selected real object.

Abstract: Embodiments are directed to neural network processing for multi-object three-dimensional (3D) modeling. An embodiment of a computer-readable storage medium includes executable computer program instructions for obtaining data from multiple cameras, the data including multiple images, and generating a 3D model for 3D imaging based at least in part on the data from the cameras, wherein generating the 3D model includes one or more of performing processing with a first neural network to determine temporal direction based at least in part on motion of one or more objects identified in an image of the multiple images or performing processing with a second neural network to determine semantic content information for an image of the multiple images.

Abstract: Methods, apparatus, systems, and articles of manufacture for real-time interactive anamorphosis projection via face detection and tracking are disclosed. An example system includes a sensor to capture an image of a face of a user. An augmented reality controller is to access the image from the sensor, determine a position of the face of the user relative to a display surface, and apply a perspective correction to an anamorphic camera representing a vantage point of the active user. A user application is to generate a scene based on the position of the anamorphic camera. A display is to present, at the display surface, the scene based on the vantage point of the active user.

Abstract: Embodiments are directed to neural network processing for multi-object three-dimensional (3D) modeling. An embodiment of a computer-readable storage medium includes executable computer program instructions for obtaining data from multiple cameras, the data including multiple images, and generating a 3D model for 3D imaging based at least in part on the data from the cameras, wherein generating the 3D model includes one or more of performing processing with a first neural network to determine temporal direction based at least in part on motion of one or more objects identified in an image of the multiple images or performing processing with a second neural network to determine semantic content information for an image of the multiple images.

Abstract: Private delivery drones and methods are disclosed. An example drone includes a first communication interface to receive a first input from a sender representing a delivery area for a payload, a second communication interface to receive a second input from a recipient representing a visual marker of the recipient, the visual marker unknown to the sender, a drone controller to, when the drone reaches the delivery area, visually identify a location in the delivery area to deliver the payload based on the visual marker, and a carrier to deliver the payload to the location.

Abstract: Technologies for detecting interactions with surfaces from a spherical view of a room include a compute device. The compute device includes an image capture manager to obtain one or more images that depict a spherical view of a room that includes multiple surfaces. Additionally, the compute device includes a surface interaction detection manager to detect, from the one or more images, a person in the room, generate a bounding box around the person, preprocess the bounding box to represent the person in an upright orientation, determine a pose of the person from the preprocessed bounding box, detect an outstretched hand from the determined pose, and determine, from the detected outstretched hand, a surface of interaction in the room.

Abstract: The present disclosure is directed to systems, apparatuses, and processes that provide mixed reality and/or augmented reality interactive environments. Disclosed embodiments include mechanisms to determine a location of a physical object within a mixed reality environment, determine a location of a viewer within the mixed reality environment, and project a display onto the physical object or on a portion of an area within the mixed reality environment proximate to the physical object to obscure the physical object from the viewer, based upon at least the location of the physical object with respect to the location of the viewer. Other embodiments may be disclosed and/or claimed.

Abstract: One embodiment provides for a graphics processor comprising a block of graphics compute units, a graphics processor pipeline coupled to the block of graphics compute units, and a programmable neural network unit including one or more neural network hardware blocks. The programmable neural network unit is coupled with the block of graphics compute units and the graphics processor pipeline. The one or more neural network hardware blocks include hardware to perform neural network operations and activation operations for a layer of a neural network. The programmable neural network unit can configure settings of one or more hardware blocks within the graphics processor pipeline based on a machine learning model trained to optimize performance of a set of workloads.

Abstract: Embodiments are directed to artificial intelligence (AI) analysis and explanation utilizing hardware measures of attention. An embodiment of a non-transitory computer-readable storage medium has stored thereon executable computer program instructions for: monitoring one or more factors of an AI network during operation of the network, the network to receive input data and output a decision based at least in part on the input data; determining attention received by the one or more factors of the network during the operation of the network; determining one or more relationships between the attention received by the one or more factors and a decision of the network based at least in part on the monitored information; and generating an analysis of the operation of the network based at least in part on the one or more relationships between attention received by the one or more factors and the decision of the network.

Abstract: Embodiments described herein provide a method comprises constructing an application tool profile from a history of tools used by an application to create one or more documents, storing the application tool profile in a memory; and creating a customized application toolset for the application using the application tool profile. Other embodiments may be described and claimed.

Abstract: Private delivery drones and methods are disclosed. An example drone includes a first communication interface to receive a first input from a sender representing a delivery area for a payload, a second communication interface to receive a second input from a recipient representing a visual marker of the recipient, the visual marker unknown to the sender, a drone controller to, when the drone reaches the delivery area, visually identify a location in the delivery area to deliver the payload based on the visual marker, and a carrier to deliver the payload to the location.

Abstract: Embodiments are directed to neural network processing for multi-object three-dimensional (3D) modeling. An embodiment of a computer-readable storage medium includes executable computer program instructions for obtaining data from multiple cameras, the data including multiple images, and generating a 3D model for 3D imaging based at least in part on the data from the cameras, wherein generating the 3D model includes one or more of performing processing with a first neural network to determine temporal direction based at least in part on motion of one or more objects identified in an image of the multiple images or performing processing with a second neural network to determine semantic content information for an image of the multiple images.