Abstract: A method for generating artificial intelligence (AI) models comprises receiving a plurality of AI models, encoding the AI models to generate a population of respective encoded hierarchical numerical representations that each encode computational units of the respective AI model as linear input-varying systems (LIVs) across a backbone, operator structure, and featurization hierarchical level, scoring each of the plurality of AI models to generate one or more respective score metrics, selecting a plurality of pairs of AI models from the received AI models based at least in part on the score metrics, for each pair, generating a respective offspring encoded hierarchical numerical representation, by combining numerical information from the encoded hierarchical numerical representations for each AI model in the respective pair, and for one or more of the offspring encoded hierarchical numerical representations, generating a respective offspring AI model based on the offspring encoded hierarchical numerical represen

Abstract: Disclosed techniques include receiving a user model as a computational graph and a list of risk metrics indicating types of risk factors to assess, and selecting at least one composable wrapper for wrapping the user model based on a risk metric(s) in the list. The techniques further include determining a set of graph-level transformations to apply to the computational graph representation that implement a risk estimate for the user model, applying the set of graph-level transformations to modify the operations performed by the user model, and generating, based on the transformed computational graph, a modified executable variant of the user model that, when executed within a machine learning framework, produces: outputs that preserve the structure and interpretation of the user model outputs; and corresponding uncertainty estimates.

Abstract: Described herein are systems and methods for improving accuracy of model output generation. A method can include obtaining a risk-aware model and a user input, applying the risk-aware model to the user input, receiving, based on the applying, model output and corresponding risk values, comparing the corresponding risk values to a threshold risk value, and regenerating the user input based on the comparing. The method can also include iteratively performing the applying, receiving, comparing, and regenerating using the regenerated user input until one or more processing conditions is met. The user input can be regenerated in response to determining that the corresponding risk values are greater than the threshold risk value. The model output can include one or more sequences in a train-of-thought (TOT) of the risk-aware model.

Abstract: Disclosed are techniques, and related systems, for developing and implementing a unified framework for quantifying risk in machine learning models, such as deep neural networks. The framework can be easy-to-use and flexible to apply to different types of machine learning models. The framework can be used to automatically assess different forms of risk in parallel, including but not limited to aleatoric uncertainty, epistemic uncertainty, and/or vacuitic uncertainty. To that end, the disclosed framework provides wrappers that compose different automatic risk assessment algorithms. The obtained risk estimates can be used, for example, for: providing a deeper insight into decision boundaries of neural networks; performing downstream tasks by integrating the risk estimates back into a learning lifecycle for the model to improve robustness and generalization; and/or improving safety by identifying potential model failures based on the risk values.

Abstract: The present disclosure provides systems and methods to detect occluded objects using shadow information to anticipate moving obstacles that are occluded behind a corner or other obstacle. The system may perform a dynamic threshold analysis on enhanced images allowing the detection of even weakly visible shadows. The system may classify an image sequence as either “dynamic” or “static”, enabling an autonomous vehicle, or other moving platform, to react and respond to a moving, yet occluded object by slowing down or stopping.

Abstract: A method of non-line-of-sight (NLoS) obstacle detection for an ego vehicle is described. The method includes capturing a sequence of images over a period with an image capture device. The method also includes storing the sequence of images in a cyclic buffer. The method further includes registering each image in the cyclic buffer to a projected image. The method includes performing the registering by estimating a homography H for each frame of the sequence of images to project to a view point of a first frame in the sequence of images and remove motion of the ego vehicle in the projected image. The method also includes enhancing the projected image. The method further includes classifying the projected image based on a scene determination. The method also includes issuing a control signal to the vehicle upon classifying the projected image.

Abstract: An object detection method includes receiving sensor data including a number of images associated with a sensor region as the actor traverses an environment, the plurality of images characterizing changes of illumination in the sensor region over time, the sensor region including a region to be traversed by the actor in the future, processing the plurality of images determine a change of illumination in sensor the region over time. The processing includes registering the plurality of images to a common coordinate system based at least in part on odometry data characterizing the actor's traversal of the environment, determining the change of illumination in the sensor region over time based on the registered plurality of images. The method further includes determining an object detection result based at least in part on the change of illumination in the sensor region over time.

Abstract: A method of non-line-of-sight (NLoS) obstacle detection for an ego vehicle is described. The method includes capturing a sequence of images over a period with an image capture device. The method also includes storing the sequence of images in a cyclic buffer. The method further includes registering each image in the cyclic buffer to a projected image. The method includes performing the registering by estimating a homography H for each frame of the sequence of images to project to a view point of a first frame in the sequence of images and remove motion of the ego vehicle in the projected image. The method also includes enhancing the projected image. The method further includes classifying the projected image based on a scene determination. The method also includes issuing a control signal to the vehicle upon classifying the projected image.

Abstract: The present disclosure provides systems and methods to detect occluded objects using shadow information to anticipate moving obstacles that are occluded behind a corner or other obstacle. The system may perform a dynamic threshold analysis on enhanced images allowing the detection of even weakly visible shadows. The system may classify an image sequence as either “dynamic” or “static”, enabling an autonomous vehicle, or other moving platform, to react and respond to a moving, yet occluded object by slowing down or stopping.

Abstract: A game ball delivery apparatus that adjusts the delivery of game balls to the actual practice needs of the player returning the delivered game balls, based on a computer analysis of player performance data, including but not limited to data collected from inertial and physiological sensors. The parameters of delivery, such as trajectory, speed, and interval of successive balls, are modified in real-time, as the player is training. One or more highly miniaturized sensors for collecting performance data are worn by the player without affecting the player's motions. A computer capable of analyzing said sensor data, selects delivery parameters for subsequently delivered balls, and transmits said delivery parameters to the ball ejector controller. Further, the system may optionally include a means for generating a feedback signal to the player to indicate deviations of the measured motion of the player from reference performances of the motion.

Abstract: A game ball delivery apparatus that adjusts the delivery of game balls to the actual practice needs of the player returning the delivered game balls, based on a computer analysis of player performance data, including but not limited to data collected from inertial and physiological sensors. The parameters of delivery, such as trajectory, speed, and interval of successive balls, are modified in real-time, as the player is training. One or more highly miniaturized sensors for collecting performance data are worn by the player without affecting the player's motions. A computer capable of analyzing said sensor data, selects delivery parameters for subsequently delivered balls, and transmits said delivery parameters to the ball ejector controller. Further, the system may optionally include a means for generating a feedback signal to the player to indicate deviations of the measured motion of the player from reference performances of the motion.

Abstract: A physical skills training system in which motion sensor data is collected and analyzed under dynamic training conditions, for the purpose of automatically identifying the type of motion being attempted, assessing said motion, and providing real-time feedback to the user so the user may adjust their subsequent orientations and movements without interruption of normal game play. The system operates under ongoing, dynamic training sessions wherein the user may continuously move about a field of play and select from a variety of positions and motion types, in addition to stationary practice scenarios. The system provides a calibration mode, in which a user may create optimum, user-specific reference profiles, which are stored in the computer, and a measured play mode in which user movements and orientations are continuously assessed against the reference motions and feedback is provided in real-time.