Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to train a neural network included in a memory augmented neural network based on one or more images and corresponding ground truth in a training dataset by transforming the one or more images to generate a plurality of one-hundred or more variations of the one or more images including variations in the ground truth and process the variations of the one or more images and store feature points corresponding to each variation of the one or more images in memory associated with the memory augmented neural network. The instructions can include further instructions to process an image acquired by a vehicle sensor with the memory augmented neural network, including comparing a feature variance set for the image acquired by the vehicle sensor to the stored processing parameters for each variation of the one or more images, to obtain an output result.

Abstract: A location of a first object can be determined in an image. A line can be drawn on the first image based on the location of the first object. A deep neural network can be trained to determine a relative location between the first object in the image and a second object in the image based on the line. The deep neural network can be optimized by determining a fitness score that divides a number of deep neural network parameters by a performance score. The deep neural network can be output.

Abstract: An end-to-end deep-learning-based system that can solve both ASR and TTS problems jointly using unpaired text and audio samples is disclosed herein. An adversarially-trained approach is used to generate a more robust independent TTS neural network and an ASR neural network that can be deployed individually or simultaneously. The process for training the neural networks includes generating an audio sample from a text sample using the TTS neural network, then feeding the generated audio sample into the ASR neural network to regenerate the text. The difference between the regenerated text and the original text is used as a first loss for training the neural networks. A similar process is used for an audio sample. The difference between the regenerated audio and the original audio is used as a second loss. Text and audio discriminators are similarly used on the output of the neural network to generate additional losses for training.

Abstract: A speech conversion system is described that includes a hierarchical encoder and a decoder. The system may comprise a processor and memory storing instructions executable by the processor. The instructions may comprise to: using a second recurrent neural network (RNN) (GRU1) and a first set of encoder vectors derived from a spectrogram as input to the second RNN, determine a second concatenated sequence; determine a second set of encoder vectors by doubling a stack height and halving a length of the second concatenated sequence; using the second set of encoder vectors, determine a third set of encoder vectors; and decode the third set of encoder vectors using an attention block.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to generate a first color image of a road environment, determine one or more value decompositions of one or more of the red, green, and blue channels of the first color image, obtain one or more modified singular value decompositions by modifying respective ones of the singular value decompositions by a non-linear equation and reconstruct a second color image based on the modified one or more singular value decompositions. The instructions can include further instructions to train a deep neural network based on the second color image and operate a vehicle based on the deep neural network.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to train a neural network included in a memory augmented neural network based on one or more images and corresponding ground truth in a training dataset by transforming the one or more images to generate a plurality of one-hundred or more variations of the one or more images including variations in the ground truth and process the variations of the one or more images and store feature points corresponding to each variation of the one or more images in memory associated with the memory augmented neural network. The instructions can include further instructions to process an image acquired by a vehicle sensor with the memory augmented neural network, including comparing a feature variance set for the image acquired by the vehicle sensor to the stored processing parameters for each variation of the one or more images, to obtain an output result.

Abstract: Example natural speech data generation systems and methods are described. In one implementation, a natural speech data generator initiates a game between a first player and a second player and determines a scenario associated with the game. A first role is assigned to the first player and a second role is assigned to the second player. The natural speech data generator receives multiple natural speech utterances by the first player and the second player during the game.

Abstract: Provided is a lab clearinghouse device configured to provide centralized management of medical tests across a plurality of medical providers, a plurality of lab payers, and a plurality of laboratories. The lab clearinghouse device is configured to communicate with a plurality of medical laboratories, medical providers, and lab payers to efficiently and effectively order and manage medical testing.

Abstract: An end-to-end deep-learning-based system that can solve both ASR and TTS problems jointly using unpaired text and audio samples is disclosed herein. An adversarially-trained approach is used to generate a more robust independent TTS neural network and an ASR neural network that can be deployed individually or simultaneously. The process for training the neural networks includes generating an audio sample from a text sample using the TTS neural network, then feeding the generated audio sample into the ASR neural network to regenerate the text. The difference between the regenerated text and the original text is used as a first loss for training the neural networks. A similar process is used for an audio sample. The difference between the regenerated audio and the original audio is used as a second loss. Text and audio discriminators are similarly used on the output of the neural network to generate additional losses for training.

Abstract: A computer includes a processor and a memory storing instructions executable by the processor to receive an image including a physical landmark, output a plurality of synthetic images, wherein each synthetic image is generated by simulating at least one ambient feature in the received image, generate respective feature vectors for each of the plurality of synthetic images, and actuate one or more vehicle components upon identifying the physical landmark in a second received image based on a similarity measure between the feature vectors of the synthetic images and a feature vector of the second received image, the similarity measure being one of a probability distribution difference or a statistical distance.

Abstract: Example natural speech data generation systems and methods are described. In one implementation, a natural speech data generator initiates a game between a first player and a second player and determines a scenario associated with the game. A first role is assigned to the first player and a second role is assigned to the second player. The natural speech data generator receives multiple natural speech utterances by the first player and the second player during the game.

Abstract: Provided is a lab clearinghouse device configured to efficiently and effectively approve medical tests across a plurality of medical laboratories, medical providers, and lab payers. The lab clearinghouse device is configured to communicate with a plurality of medical laboratories, medical providers, and lab payers to efficiently and effectively approve, direct, and manage medical tests.

Abstract: A computing system can determine a vehicle command based on a received spoken language command and determined confidence levels. The computing system can operate a vehicle based on the vehicle command. The computing system can further determine the spoken language command by processing audio spectrum data corresponding to spoken natural language with an automatic speech recognition (ASR) system.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to generate a first color image of a road environment, determine one or more value decompositions of one or more of the red, green, and blue channels of the first color image, obtain one or more modified singular value decompositions by modifying respective ones of the singular value decompositions by a non-linear equation and reconstruct a second color image based on the modified one or more singular value decompositions. The instructions can include further instructions to train a deep neural network based on the second color image and operate a vehicle based on the deep neural network.

Abstract: Systems, methods, and devices for speech transformation and generating synthetic speech using deep generative models are disclosed. A method of the disclosure includes receiving input audio data comprising a plurality of iterations of a speech utterance from a plurality of speakers. The method includes generating an input spectrogram based on the input audio data and transmitting the input spectrogram to a neural network configured to generate an output spectrogram. The method includes receiving the output spectrogram from the neural network and, based on the output spectrogram, generating synthetic audio data comprising the speech utterance.

Abstract: Provided is a lab clearinghouse device configured to reduce fraud, waste, and abuse in the ordering and performance of medical testing. The lab clearinghouse device is configured to communicate with a plurality of medical laboratories, medical providers, and lab payers to efficiently and effectively reduce fraud, waste, and abuse in the ordering and performance of medical testing.

Abstract: A computing system can translate a spoken natural language command into an intermediate constructed language command with a first deep neural network and determine a vehicle command and an intermediate constructed language response with a second deep neural network based on receiving vehicle information. The computing system can translate the intermediate constructed language response into a spoken natural language response with a third deep neural network and operate a vehicle based on the vehicle command.

Abstract: A computing system can be programmed to receive a spoken language command in response to emitting a spoken language cue and process the spoken language command with a generalized adversarial neural network (GAN) to determine a vehicle command. The computing system can be further programmed to operate a vehicle based on the vehicle command.

Abstract: A speech conversion system is described that includes a hierarchical encoder and a decoder. The system may comprise a processor and memory storing instructions executable by the processor. The instructions may comprise to: using a second recurrent neural network (RNN) (GRU1) and a first set of encoder vectors derived from a spectrogram as input to the second RNN, determine a second concatenated sequence; determine a second set of encoder vectors by doubling a stack height and halving a length of the second concatenated sequence; using the second set of encoder vectors, determine a third set of encoder vectors; and decode the third set of encoder vectors using an attention block.

Abstract: A method for controlling an autonomous vehicle is disclosed. The method may include receiving image data. The image data may be logistically classified by a first neural network as pertaining to one situation of a plurality of situations. Based on this logistic classification, the image data may be assigned to a second neural network that is specifically trained to address the one situation. The second neural network may perform regression on the image data. Thereafter, the vehicle may be control with a command based on the regression.