Abstract: The present disclosure relates to a computer-implemented method of training a generative model to insert an object in spatial sensor data. The method comprises receiving a training sample of spatial sensor data and receiving an indication of a 3D geometric property of an object captured in the training sample. A portion of spatial sensor data corresponding to the object is removed from the training sample, resulting a cropped training sample. The generative model is trained to reconstruct the training sample from the cropped training sample by, providing to the generative model: the cropped training sample as a target input, an indication of the object as a reference input, and the 3D geometric property of the object as a conditioning input. This results in a generated output sample of spatial sensor data.

Abstract: The present disclosure relates to techniques for training a generative model to insert an object in spatial sensor data. A first training sample of spatial sensor data of a first sensor modality, and a second training sample of spatial sensor data of a second sensor modality are received, the first training sample and the second training sample capture a common object. A first portion of sensor data corresponding to the object is removed from the first training sample, resulting a cropped training sample. A second portion of spatial sensor data corresponding to the common object is extracted from the second training sample.

Abstract: A computer-implemented method of modelling a perception system for perceiving objects captured in sensor data comprises: receiving a plurality of training examples, each comprising a ground truth scene for a set of sensor data and a corresponding perceived scene obtained by applying the perception system to the set of sensor data; fitting to the training examples noise model parameters, encoding a noise distribution over perceived scenes given a misdetection scene, and misdetection model parameters, encoding a misdetection distribution over misdetection scenes given a ground truth scene; computing a perception distribution over perceived scenes for a given ground truth scene by marginalizing the product of noise and misdetection distributions over multiple misdetection scenes, wherein individual objects in the ground truth scene are not associated with individual objects in the perceived scenes; fitting the noise and misdetection model parameters to match the perception distribution to the perceived scene for

Abstract: A computer-implemented method of generating black-box adversarial inputs to a perception component using a surrogate model of the perception component comprises receiving an initial input to the perception component and repeatedly perturbing the initial input until an adversarial input is found that satisfies an attack objective by: performing a primary attack process by perturbing the initial input based on a computed gradient of a surrogate attack loss function of the surrogate model that encodes the attack objective; wherein, if the primary attack process terminates without finding any perturbed input satisfying the promising attack condition, a backup attack process is performed to perform a randomized search of the input space of the perception component, guided by the surrogate model, until a perturbed input satisfying the promising attack condition is found; wherein the primary attack process is repeated based on the perturbed input found by the primary attack process or backup process.

Abstract: A computer-implemented method of generating black-box adversarial inputs to a perception component comprises computing an adversarial input by applying a perturbation to an original input, the adversarial input satisfying an attack objective when inputted to the perception component. The perturbation is determined by selectively combining component perturbations selected from a predetermined set of component perturbations. Inputs correspond to respective points in an input vector space, and the component perturbations encode principal attack directions in the input vector space for satisfying said attack objective, the principal attack directions having been determined by analyzing: (i) a set of sample attack directions, or (ii) a set of input samples.

Abstract: A perception model is trained to classify inputs in relation to a discrete set of leaf node classes. A hierarchical classification tree encodes hierarchical relationships between the leaf node classes. A training loss function is dependent on a classification score for a given training input a its ground truth leaf node class of the training input, but also classification scores for at least some others of the leaf node classes, with the classification scores of the other leaf node classes weighted in dependence on their hierarchical relationship to the ground truth leaf node class within the hierarchical classification tree.

Abstract: A computer-implemented method of modelling a perception system for perceiving objects captured in sensor data comprises: receiving a plurality of training examples, each comprising a ground truth scene for a set of sensor data and a corresponding perceived scene obtained by applying the perception system to the set of sensor data; fitting to the training examples noise model parameters, encoding a noise distribution over perceived scenes given a misdetection scene, and misdetection model parameters, encoding a misdetection distribution over misdetection scenes given a ground truth scene; computing a perception distribution over perceived scenes for a given ground truth scene by marginalizing the product of noise and misdetection distributions over multiple misdetection scenes, wherein individual objects in the ground truth scene are not associated with individual objects in the perceived scenes; fitting the noise and misdetection model parameters to match the perception distribution to the perceived scene for

Abstract: A perception model is trained to classify inputs in relation to a discrete set of leaf node classes. A hierarchical classification tree encodes hierarchical relationships between the leaf node classes. A training loss function is dependent on a classification score for a given training input a its ground truth leaf node class of the training input, but also classification scores for at least some others of the leaf node classes, with the classification scores of the other leaf node classes weighted in dependence on their hierarchical relationship to the ground truth leaf node class within the hierarchical classification tree.