Abstract: A method for managing memory access for implementing at least one layer of a convolutional neural network is provided. The method comprises predicting an access procedure in relation to a portion of memory based on a characteristic of the convolutional neural network. In response to the prediction, the method comprises performing an operation to obtain and store a memory address translation, corresponding to the portion of memory, in storage in advance of the predicted access procedure. An apparatus is provided comprising at least one processor and storage. The apparatus is configured to predict an access procedure in relation to a portion of memory which is external to the processor. In response to the prediction, the apparatus is configured to obtain and store a memory address translation corresponding to the portion of memory in storage in advance of the predicted access procedure.

Abstract: When rendering a new frame using a hybrid ray tracing process, a graphics processor transforms vertex position data for models representing objects in the new frame into camera space for use in a rasterisation process of the hybrid ray tracing process irrespective of whether the object has moved between frames or whether the geometry of the model representing object has changed, but only transforms vertex position data for a model for an object into world space for use in a ray tracing process of the hybrid ray tracing process when the object has moved between frames or where the geometry of the model representing the object has changed.

Abstract: A method of data processing for an augmented reality system. The method comprises obtaining augmented reality data output by an augmented reality application operating at a second trust level. The augmented reality data is for modifying a representation of a real-world environment for a user of the augmented reality system. The method also comprises obtaining object recognition data determined by an object recognition system operating at a first trust level. The object recognition data comprises an indication of an object belonging to a predetermined class of objects being present in the real-world environment. The method also comprises triggering modification of the augmented reality data in response to the object recognition data, based on prioritization of the first trust level over the second trust level.

Abstract: Examples of the present disclosure relate to methods for performing object detection. In one such example, data representing an image is received. The image comprises at least one target region and a further region. The at least one target region is identifiable using data indicative of a gaze direction of a viewer of the image. A first portion of the data is processed using a first processing scheme to perform object detection in the at least one target region of the image. The first portion of the data represents the at least one target region of the image. A second portion of the data is processed using a second, different, processing scheme. The second portion of the data represents the further region of the image.

Abstract: When rendering a new frame using a hybrid ray tracing process, a graphics processor transforms vertex position data for models representing objects in the new frame into camera space for use in a rasterisation process of the hybrid ray tracing process irrespective of whether the object has moved between frames or whether the geometry of the model representing object has changed, but only transforms vertex position data for a model for an object into world space for use in a ray tracing process of the hybrid ray tracing process when the object has moved between frames or where the geometry of the model representing the object has changed.

Abstract: A computer implemented method for performing convolutions between subsets of an input data array and a kernel resulting in subsets of an output data array. The method may include receiving an input data array and using positional data indicating the position of elements of the input data array to determine subsets of the input data array which contains at least one non-zero value data element; performing convolutions between the subsets of the input data array containing at least one non-zero value data element and a kernel to produce output data array subsets; and combining the output data subsets with the positional data to generate output data indicative of a completed output data array.

Abstract: A method comprising: obtaining pose data representative of a pose of a portable device during observation of an environment comprising an object; obtaining distance data representative of a distance between the object and a receiver during the observation of the environment, using at least one radio waveform reflected from the object and received by the receiver; and processing the pose data and the distance data to generate a topological model of the object.

Abstract: In a data processing system, when displaying frames to a user on a display, the location of the user's gaze on the display is determined, and the amount of light to be output for different regions of the display when displaying the frame on the display is selected and set based on the determined location of the user's gaze on the display.

Abstract: A method of processing image data representative of an image using a multi-stage system comprising a first neural network (NN) for identifying a first image characteristic and a second NN for identifying a second image characteristic. The method comprises processing the image data using t a first at least one layer of the first NN to generate feature data representative of at least one feature of the image and processing the feature data using a second at least one layer of the first NN to generate first image characteristic data indicative of whether the image includes the first image characteristic. The feature data is transferred from the first NN to the second NN. The feature data is processed using the second NN to generate second image characteristic data representative of whether the image includes the second image characteristic.

Abstract: A method for processing video data, comprising: receiving a stream of input video data representative of a number of successive frames generated by an imaging device's image sensor; selecting at least some of the frames; for each selected frame: determining, using an image signature algorithm, a signature for each region of the given selected frame; and, based on such signatures, classifying each region in that frame as either a changing or a static region; and generating an output video data stream that is a compressed version of the input video data, with a greater average data reduction rate for static region data than for changing region data, of the selected frames. The signature algorithm is such that a region's signature has substantially smaller size than the input data representative of that region, and such that signatures for visually similar regions are the same or similar.

Abstract: Examples of the present disclosure relate to methods for processing image data. In one such example, first data representing a rendered image is received. In some cases, second data useable to identify at least one target region of the rendered image is received, the at least one target region being associated with a gaze direction of a viewer. A first portion of the first data is processed in accordance with a first data reduction rate to derive first processed data, the first portion representing the at least one target region. A second portion of the first data is processed in accordance with a second data reduction rate, different from the first data reduction rate. The second portion represents a further region of the rendered image, different from the at least one target image. At least the first processed data is outputted.

Abstract: A method of super-resolution image processing. The method includes inputting first image data representative of a first version of at least part of an image with a first resolution to a machine learning system. The first image data includes pixel intensity data representative of an intensity value of at least one color channel of a pixel of the first version of the at least part of the image, and feature data representative of a value of at least one non-intensity feature associated with the pixel. The first image data is processed using the machine learning system to generate second image data representative of a second version of the at least part of the image with a second resolution greater than the first resolution.

Abstract: A navigation system uses markers that are identifiable in images of an environment being navigated to determine the location of a portable device in the environment. The portable device takes images of the environment, and those images are analysed to identify markers in the images and the pose of the portable device based on the image of the marker. The identified marker and the determined pose of the portable device are then used to determine the location and orientation of the portable device in the environment being navigated.

Abstract: A method of operating a data processing system is disclosed for a data processing system that comprises a display and a display controller operable to provide to the display data in respect of output surfaces to be displayed. The method comprises, when an output surface is to be displayed, the display controller providing to the display data in respect of the output surface in the form of image data and image modification data, and the display using the image data and the image modification data when producing an output surface for display.

Abstract: When generating an extrapolated frame by extrapolating object motion from a rendered frame in a graphics processing system, regions of a first frame are compared 90 with corresponding regions of another frame. If the regions are determined to be different, it is determined whether or not regions of the first frame contain objects that need to be extrapolated to generate the extrapolated frame, but if the regions are determined to be similar, it is assumed that the regions of the first frame do not contain any objects that need to be extrapolated to generate the extrapolated frame.

Abstract: Examples of the present disclosure relate to methods for controlling a display device. In one such example, data representing a result of an eyewear detection operation is obtained. Dependent on the obtained data, a control signal is outputted to adjust a display parameter of the display device. Performing the eyewear detection operation comprises receiving image data representing a user of the display device, and processing the image data using object recognition to determine whether or not the user is wearing eyewear of a predetermined type.

Abstract: A method of operating a graphics processing system that generates “spacewarped” frames for display is disclosed. Primitive motion vectors are used to determine the motion of objects appearing in rendered application frames. The so-determined motion is then used to generate “spacewarped” versions of the rendered application frames.

Abstract: A method of processing video data representative of a video comprising a first and second frame to generate output data representative of at least one feature of the second frame. The method includes identifying a first and second portion of the second frame, which correspond to a first and second portion of the first frame, respectively. First feature data obtained by processing first frame data associated with the first portion of the first frame using a first feature detection operation is retrieved from storage. Second feature data representative of a second feature map is generated by processing second frame data associated with the second portion of the second frame using the first feature detection operation. The first feature data and the second feature data are processed using a second feature detection operation to generate the output data.

Abstract: A method for managing memory access for implementing at least one layer of a convolutional neural network is provided. The method comprises predicting an access procedure in relation to a portion of memory based on a characteristic of the convolutional neural network. In response to the prediction, the method comprises performing an operation to obtain and store a memory address translation, corresponding to the portion of memory, in storage in advance of the predicted access procedure. An apparatus is provided comprising at least one processor and storage. The apparatus is configured to predict an access procedure in relation to a portion of memory which is external to the processor. In response to the prediction, the apparatus is configured to obtain and store a memory address translation corresponding to the portion of memory in storage in advance of the predicted access procedure.

Abstract: A method for producing a context-sensitive augmented reality output comprising the steps of obtaining an output of an image sensor, wherein the output is representative of a real-world environment. Environmental data associated with the real-world environment is then obtained and analyzed to determine at least one environmental characteristic. At least one virtual object is then modified based on the at least one environmental characteristic derived from the environmental data. The modified virtual object is then inserted into the output of the image sensor to produce a modified output; which is then displaying on a display.