Abstract: A method for computing classifications of raw tomographic data includes: supplying the raw tomographic data to a sinogram-convolutional neural network including blocks, at least one of the blocks being configured to perform a convolution of the raw tomographic data in Radon space with a convolutional kernel by: slicing the raw tomographic data into a plurality of one-dimensional tomographic data slices along an angle dimension of the raw tomographic data; slicing the convolutional kernel into a plurality of one-dimensional kernel slices along the angle dimension of the convolutional kernel; for each angle, computing a one-dimensional convolution between: a corresponding one of the one-dimensional tomographic data slices at the angle; and a corresponding one of the one-dimensional kernel slices at the angle; and collecting the one-dimensional convolutions at the angles; computing a plurality of features from the convolution; and computing the classifications of the raw tomographic data based on the features.

Abstract: Described is a system for classifying objects and scenes in images. The system identifies salient regions of an image based on activation patterns of a convolutional neural network (CNN). Multi-scale features for the salient regions are generated by probing the activation patterns of the CNN at different layers. Using an unsupervised clustering technique, the multi-scale features are clustered to identify key attributes captured by the CNN. The system maps from a histogram of the key attributes onto probabilities for a set of object categories. Using the probabilities, an object or scene in the image is classified as belonging to an object category, and a vehicle component is controlled based on the object category causing the vehicle component to perform an automated action.

Abstract: Described is a stripmap SAR system on a vehicle comprising an antenna that is fixed and directed outward from the side of the vehicle, a SAR sensor, a storage, and a computing device. The computing device comprises a memory, one or more processing units, and a machine-readable medium on the memory. The machine-readable medium stores instructions that, when executed by the one or more processing units, cause the stripmap SAR system to perform various operations. The operations comprise: receiving stripmap range profile data associated with observed views of a scene; transforming the received stripmap range profile data into partial circular range profile data; comparing the partial circular range profile data to a template range profile data of the scene; and estimating registration parameters associated with the partial circular range profile data relative to the template range profile data to determine a deviation from the template range profile data.

Abstract: Described is a system for learning actions for image-based action recognition in an autonomous vehicle. The system separates a set of labeled action image data from a source domain into components. The components are mapped onto a set of action patterns, thereby creating a dictionary of action patterns. For each action in the set of labeled action data, a mapping is learned from the action pattern representing the action onto a class label for the action. The system then maps a set of new unlabeled target action image data onto a shared embedding feature space in which action patterns can be discriminated. For each target action in the set of new unlabeled target action image data, a class label for the target action is identified. Based on the identified class label, the autonomous vehicle is caused to perform a vehicle maneuver corresponding to the identified class label.

Abstract: Described is a system for object recognition. The system generates a training image set of object images from multiple image classes. Using a training image set and annotated semantic attributes, a model is trained that maps visual features from known images to the annotated semantic attributes using joint sparse representations with respect to dictionaries of visual features and semantic attributes. The trained model is used for mapping visual features of an unseen input image to its semantic attributes. The unseen input image is classified as belonging to an image class, and a device is controlled based on the classification of the unseen input image.

Abstract: Described is a system for continual learning using experience replay. In operation, the system receives a plurality of tasks sequentially, from which a current task is fed to an encoder. The current task has data points associated with the current task. The encoder then maps the data points into an embedding space, which reflects the data points as discriminative features. A decoder then generates pseudo-data points from the discriminative features, which are provided back to the encoder. The discriminative features are updated in the embedding space based on the pseudo-data points. The encoder then learns (updates) a classification of a new task by matching the new task with the discriminative features in the embedding space.

Abstract: Described is a system for detecting backdoor attacks in deep convolutional neural networks (CNNs). The system compiles specifications of a pretrained CNN into an executable model, resulting in a compiled model. A set of Universal Litmus Patterns (ULPs) are fed through the compiled model, resulting in a set of model outputs. The set of model outputs are classified and used to determine presence of a backdoor attack in the pretrained CNN. The system performs a response based on the presence of the backdoor attack.

Abstract: Described is a system for detecting change of context in a video stream on an autonomous platform. The system extracts salient patches from image frames in the video stream. Each salient patch is translated to a concept vector. A recurrent neural network is enervated with the concept vector, resulting in activations of the recurrent neural network. The activations are classified, and the classified activations are mapped onto context classes. A change in context class is detected in the image frames, and the system causes the autonomous platform to perform an automatic operation to adapt to the change of context class.

Abstract: Systems and methods according to one or more embodiments are provided for registration of synthetic aperture range profile data to aid in SAR-based navigation. In one example, a SAR-based navigation system includes a memory comprising a plurality of executable instructions. The SAR-based navigation system further includes a processor adapted to receive range profile data associated with observed views of a scene, compare the range profile data to a template range profile data of the scene, and estimate registration parameters associated with the range profile data relative to the template range profile data to determine a deviation from the template range profile data.

Abstract: Described is a system for understanding machine-learning decisions. In an unsupervised learning phase, the system extracts, from input data, concepts represented by a machine-learning (ML) model in an unsupervised manner by clustering patterns of activity of latent variables of the concepts, where the latent variables are hidden variables of the ML model. The extracted concepts are organized into a concept network by learning functional semantics among the extracted concepts. In an operational phase, a subnetwork of the concept network is generated. Nodes of the subnetwork are displayed as a set of visual images that are annotated by weights and labels, and the ML model per the weights and labels.

Abstract: Described is a system that can recognize novel objects that the system has never before seen. The system uses a training image set to learn a model that maps visual features from known images to semantic attributes. The learned model is used to map visual features of an unseen input image to semantic attributes. The unseen input image is classified as belonging to an image class with a class label. A device is controlled based on the class label.

Abstract: Described is a system for predicting multi-agent movements. A Radon Cumulative Distribution Transform (Radon-CDT) is applied to pairs of signature-formations representing agent movements. Canonical correlation analysis (CCA) components are identified for the pairs of signature-formations. Then, a relationship between the pairs of signature formations is learned using the CCA components. A counter signature-formation for a new dataset is predicted using the learned relationship and a new signature-formation. Control parameters of a device can be adjusted based on the predicted counter signature-formation.

Abstract: Described is a system for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain. The system uses a measured similarity between the EO domain and the SAR domain to train a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain. Using the trained model, a SAR image is processed to determine regions of interest in the SAR image. A region of interest is classified to determine whether the region of interest corresponds to an object of interest, and classified regions of interest that contain the object of interest are output. The object of interest is displayed on a visualization map, and the visualization map is automatically updated to reflect a change in position of the object of interest.

Abstract: Described is a system for discriminant localization of objects. During operation, the system causes one or more processors to perform an operation of identifying an object in an image using a multi-layer network. Features of the object are derived from the activations of two or more layers of the multi-layer network. The image is then classified to contain one or more object classes, and the desired object class is localized. A device can then be controlled based on localization of the object in the image. For example, a robotic arm can be controlled to reach for the object.

Abstract: A method for training a controller to control a robotic system includes: receiving a neural network of an original controller for the robotic system based on origin data samples from an origin domain and labels in a label space, the neural network including encoder and classifier parameters, the neural network being trained to: map an input data sample from the origin domain to a feature vector in a feature space using the encoder parameters; and assign a label of the label space to the input data sample using the feature vector based on the classifier parameters; updating the encoder parameters to minimize a dissimilarity, in the feature space, between: origin feature vectors computed from the origin data samples; and target feature vectors computed from target data samples from a target domain; and updating the controller with the updated encoder parameters to control the robotic system in the target domain.

Abstract: An autonomous or semi-autonomous system includes a temporal prediction network configured to process a first set of samples from an environment of the system during performance of a first task, a controller configured to process the first set of samples from the environment and a hidden state output by the temporal prediction network, a preserved copy of the temporal prediction network, and a preserved copy of the controller. The preserved copy of the temporal prediction network and the preserved copy of the controller are configured to generate simulated rollouts, and the system is configured to interleave the simulated rollouts with a second set of samples from the environment during performance of a second task to preserve knowledge of the temporal prediction network for performing the first task.

Abstract: Described is a system for unsupervised domain adaptation in an autonomous learning agent. The system adapts a learned model with a set of unlabeled data from a target domain, resulting in an adapted model. The learned model was previously trained to perform a task using a set of labeled data from a source domain. The set of labeled data has a first input data distribution, and the set of unlabeled target data has a second input data distribution that is distinct from the first input data distribution. The adapted model is implemented in the autonomous learning agent, causing the autonomous learning agent to perform the task in the target domain.

Abstract: A method for retraining an artificial neural network trained on data from an old task includes training the artificial neural network on data from a new task different than the old task, calculating a drift, utilizing Sliced Wasserstein Distance, in activation distributions of a series of hidden layer nodes during the training of the artificial neural network with the new task, calculating a number of additional nodes to add to at least one hidden layer based on the drift in the activation distributions, resetting connection weights between input layer nodes, hidden layer nodes, and output layer nodes to values before the training of the artificial neural network on the data from the new task, adding the additional nodes to the at least one hidden layer, and training the artificial neural network on data from the new task.

Abstract: Described is a system for classifying novel objects in imagery. In operation, the system extracts salient patches from a plurality of unannotated images using a multi-layer network. Activations of the multi-layer network are clustered into key attribute, with the key attributes being displayed to a user on a display, thereby prompting the user to annotate the key attributes with class label. An attribute database is then generated based on user prompted annotations of the key attributes. A test image can then be passed through the system, allowing the system to classify at least one object in the test image by identifying an object class in the attribute database. Finally, a device can be caused to operate or maneuver based on the classification of the at least one object in the test image.

Abstract: Described is a system for prediction of adversary movements. In an aspect, the system includes one or more processors and a memory, the memory being a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions, the one or more processors perform operations of computing relative positions of multiple objects of interest, generating a feature representation by forming a matrix based on the relative positions, predicting movement of the multiple objects of interest by applying clustering to the feature representation and by performing canonical correlation analysis, and controlling a device based on the predicted movement of the multiple objects of interest.