Abstract: A method of generating a training dataset suitable for training machine learning algorithms to estimate the motion of objects, and for training a machine learning algorithm to perform motion estimation. A plurality of pairs of synthetic images are generated from obtained objects and backgrounds, each pair have a first frame and a second frame. The first frame includes a selection of objects in first positions and first orientations superimposed on a selected background, and the second frame includes the selection of objects in second positions and second orientations superimposed on the selected background. Also provided are processing systems configured to carry out these methods.

Abstract: A hardware-implemented method of indexing data elements in a source array in a memory, generates a number of shifted copy arrays based on the source array, each shifted copy array comprising the data elements of the source array at a respective shifted position. A plurality of indices for indexing the source array are received, each index of the plurality of indices indicating a target location in the source array, and for each index of the plurality of indices, a data element is retrieved from each of the shifted copy arrays. The retrieved elements are gated based on the index, to thereby select a data element.

Abstract: A windowed operation is implemented in at least three traversed dimensions. The windowed operation applies a window having at least three dimensions to data having at least three traversed dimensions, with shifts of the window in all three traversed dimensions. Two dimensions of the at least three traversed dimensions are selected, and the windowed operation is mapped to a plurality of constituent 2-D windowed operations in the selected two dimensions, the 2-D windowed operations applying a slice of the window to a slice of the data, with shifts of the slice of the window in only two dimensions. Each of the plurality of 2-D windowed operations is implemented by at least one hardware accelerator, each 2-D windowed operation producing a respective partial result, and the partial results are assembled to produce the result of the windowed operation.

Abstract: A hardware-implemented method of indexing data elements in a source array is provided. The method comprises generating a number of shifted copy arrays; receiving indices for indexing the source array; and retrieving one or more data elements from the shifted copy arrays, according to the received indices. Also disclosed is a related processing system comprising a memory and hardware for indexing data elements in a source array in the memory.

Abstract: A data processing system and method are disclosed, for implementing a windowed operation in at least three traversed dimensions. The data processing system maps the windowed operation in at least three traversed dimensions to a plurality of constituent windowed operations in two traversed dimensions. This plurality of 2-D windowed operations is implemented as such in at least one hardware accelerator. The data processing system assembles the results of the constituent 2-D windowed operations to produce the result of the windowed operation in at least three traversed dimensions.

Abstract: A method and data processing system for resampling a first set of samples using a neural network accelerator. The first set of samples is arranged in a tensor extending in at least a first dimension defined in a first coordinate system. A set of resampling parameters is determined, having a first resampling factor a_1/b_1 for a first dimension, and a first offset d_1 for the first dimension. At least a first number of kernels is obtained, and the first set of samples is resampled to produce a second set of samples, based on the first resampling factor and the first offset.

Abstract: A method of analysing objects in a first frame and a second frame is disclosed. The method includes segmenting the frames, and matching at least one object in the first frame with a corresponding object in the second frame. The method optionally includes estimating the motion of the at least one matched object between the frames. Also disclosed is a method of generating a training dataset suitable for training machine learning algorithms to estimate the motion of objects. Also provided are processing systems configured to carry out these methods.

Abstract: A method of generating a training dataset suitable for training machine learning algorithms to estimate the motion of objects, and for training a machine learning algorithm to perform motion estimation. A plurality of pairs of synthetic images are generated from obtained objects and backgrounds, each pair have a first frame and a second frame. The first frame includes a selection of objects in first positions and first orientations superimposed on a selected background, and the second frame includes the selection of objects in second positions and second orientations superimposed on the selected background. Also provided are processing systems configured to carry out these methods.

Abstract: A method of implementing a scatter operation in fixed-function hardware of a neural network accelerator involves converting two or more vectors of indices to sparse index tensors in a one-hot sparse format. An update tensor is generated, by applying the update values to one of the sparse index tensors (or a tensor derived from it). In some examples, an input data tensor is updated based on the update tensor. In other examples, the update tensor itself is output.

Abstract: A method of implementing a scatter operation in fixed-function hardware of a neural network accelerator involves converting two or more vectors of indices to sparse index tensors in a one-hot sparse format. An update tensor is generated, by applying the update values to one of the sparse index tensors (or a tensor derived from it). In some examples, an input data tensor is updated based on the update tensor. In other examples, the update tensor itself is output.

Abstract: A method of implementing a scatter operation in fixed-function hardware of a neural network accelerator involves converting two or more vectors of indices to sparse index tensors in a one-hot sparse format. An update tensor is generated, by applying the update values to one of the sparse index tensors (or a tensor derived from it). In some examples, an input data tensor is updated based on the update tensor. In other examples, the update tensor itself is output.

Abstract: Methods of dividing a neural network into chunks of operations executable in a hardware pass of hardware to execute a neural network. The layers of the neural network are divisible into layer groups that comprise a sequence of layers executable in the same hardware pass of the hardware. Each layer group is divisible into chunks of operations executable in a hardware pass of the hardware. The chunks for a layer group are defined by split parameters. A layer group loss function is obtained that represents a performance metric associated with executing a layer group on the hardware as a function of the split parameters and neural network architecture parameters for the layer group.

Abstract: A mechanism for processing, on a hardware accelerator comprising fixed-function circuitry, data according to a neural network process that includes a pooling, unpooling or backward pooling and/or binary argmax/argmin function. The function is mapped to a set of elementary neural network operations available to the fixed-function circuitry. The neural network process is then executed using the fixed-function circuitry. The data processed using the neural network process comprises image and/or audio data.

Abstract: A method and data processing system for implementing a neural network containing at least one matrix multiplication operation. The matrix multiplication operation is mapped to a graph of neural network operations including at least one element-wise operation. The at least one element-wise operation is implemented in fixed-function hardware of a neural network accelerator.

Abstract: A mechanism for processing, on a hardware accelerator comprising fixed-function circuitry, data according to a neural network process that comprises a neural network with an associated argmax or argmin function. The argmax or argmin function is mapped to a set of elementary neural network operations available to the fixed-function circuitry. The neural network process is then executed using the fixed-function circuitry. The data processed using the neural network process comprises image and/or audio data.

Abstract: Methods for implementing an exponential operation, and a softmax neural network layer, in neural network accelerator hardware, and a data processing system for implementing the exponential operation and a data processing system for implementing the softmax layer. The exponential operation or softmax layer is mapped to a plurality of elementary neural network operations, and the neural network accelerator hardware evaluates these operations, to produce the result of the operation or layer respectively.

Abstract: Methods and systems of generating gradients of a loss metric for a neural network (NN) with respect to weights of a convolution layer of the NN, the convolution layer of the NN configured to receive an input tensor of input values and a weight tensor of weights, and generate an output tensor of output values.

Abstract: A method and data processing system implement dilated convolution operations in hardware. Embodiments provide various ways to implement a dilated convolution based on a number of constituent convolutions, by either splitting the kernel to construct a set of constituent convolutions with smaller kernels, or dividing the input data into multiple parts and applying a convolution to each part separately. The constituent convolutions are evaluated in hardware and their results are combined to produce the result of the dilated convolution.

Abstract: A method and data processing system implement a neural network containing at least one matrix multiplication operation. The matrix multiplication operation is mapped to a graph of neural network operations including at least one transformation and at least one convolution. The at least one convolution is implemented in fixed-function hardware of a neural network accelerator.

Abstract: A data processing system and method are disclosed, for implementing a windowed operation in at least three traversed dimensions. The data processing system maps the windowed operation in at least three traversed dimensions to a plurality of constituent windowed operations in two traversed dimensions. This plurality of 2-D windowed operations is implemented as such in at least one hardware accelerator. The data processing system assembles the results of the constituent 2-D windowed operations to produce the result of the windowed operation in at least three traversed dimensions.