Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for accurately and efficiently learning parameters of a distilled neural network from parameters of a source neural network utilizing multiple augmentation strategies. For example, the disclosed systems can generate lightly augmented digital images and heavily augmented digital images. The disclosed systems can further learn parameters for a source neural network from the lightly augmented digital images. Moreover, the disclosed systems can learn parameters for a distilled neural network from the parameters learned for the source neural network. For example, the disclosed systems can compare classifications of heavily augmented digital images generated by the source neural network and the distilled neural network to transfer learned parameters from the source neural network to the distilled neural network via a knowledge distillation loss function.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed that improve detection of malware based on ecosystem specific data. An example apparatus includes a feedback weight controller to apply, with a machine learning model, a weight to feedback associated with a sample, the feedback obtained from at least a customer ecosystem and including endpoint feedback, human feedback, infrastructure feedback, and global feedback; and a sample conviction controller to, in response to a score based on the weighted feedback satisfying a threshold for a classification, indicate to a user, with the machine learning model, that the classification for the sample is malicious.

Abstract: Aspects of the subject disclosure may include, for example, obtaining a first plurality of inputs that identify a plurality of geographical locations and a plurality of infrastructure located at the plurality of geographical locations, classifying each of the plurality of geographical locations in accordance with the first plurality of inputs to obtain a plurality of classes, obtaining a second plurality of inputs that identify costs, revenue, profits, or any combination thereof, associated with the plurality of infrastructure, processing the second plurality of inputs in conjunction with the plurality of classes to identify a first plurality of locations included in the plurality of geographical locations to decommission infrastructure included in the plurality of infrastructure, and presenting the first plurality of locations via a device. Other embodiments are disclosed.

Abstract: In one embodiment, a device receives optical time domain reflectometer (OTDR) trace samples, each sample labeled with an associated fiber optic cable condition. The device alters the received OTDR trace samples to generate a set of synthetic OTDR trace samples. Each synthetic sample is labeled with the label of the received sample that was altered to generate the synthetic sample. The device trains a machine learning-based classifier using a training dataset that comprises the synthetic OTDR trace samples. The device uses the trained classifier to identify a condition along a particular fiber optic cable based on OTDR trace data obtained from that cable.

Abstract: Approaches are presented for training and using scene graph generators for transfer learning. A scene graph generation technique can decompose a domain gap into individual types of discrepancies, such as may relate to appearance, label, and prediction discrepancies. These discrepancies can be reduced, at least in part, by aligning the corresponding latent and output distributions using one or more gradient reversal layers (GRLs). Label discrepancies can be addressed using self-pseudo-statistics collected from target data. Pseudo statistic-based self-learning and adversarial techniques can be used to manage these discrepancies without the need for costly supervision from a real-world dataset.

Abstract: A method for extracting and providing a text color and background color in an image, includes detecting a first area that includes a text in a given image; extracting, from the first area, a representative text color that represents the text and a representative background color that represents a background of the first area; and overlaying a second area that includes a translation result of the text on the given image and applying the representative text color and the representative background color to a text color and a background color of the second area.

Abstract: A device configured for more efficiently processing video images within a set of video image data to detect objects is described herein. The device may include a processor configured to execute a neural network such as a convolutional neural network. The device can receive video image data from a plurality of cameras, such as stationary cameras. The device can acquire a set of sample images from a stationary camera and submit them to a specialized neural network for processing to generate an attention mask. The attention mask can be generated from a variety of methods and is applied to each of the subsequently acquired images form the camera to narrow down areas where the convolutional neural network should process data. The application of attention masks to images within video image data creates masked images that can be processed to detect objects with much greater accuracy and fewer computational resources required.

Abstract: Computer vision systems and methods for text classification are provided. The system detects a plurality of text regions in an image and generates a bounding box for each detected text region. The system utilizes a neural network to recognize text present within each bounding box and classifies the recognized text, based on at least one extracted feature of each bounding box and the recognized text present within each bounding box, according to a plurality of predefined tags. The system can associate a key with a value and return a key-value pair for each predefined tag.

Abstract: The present disclosure relates to systems, methods, and computer readable media that evaluate performance of a machine learning system in connection with a test dataset. For example, systems disclosed herein may receive a test dataset and identify label information for the test dataset including feature information and ground truth data. The systems disclosed herein can compare the ground truth data and outputs generated by a machine learning system to evaluate performance of the machine learning system with respect to the test dataset. The systems disclosed herein may further generate feature clusters based on failed outputs and corresponding features and generate a number of performance views that illustrate performance of the machine learning system with respect to clustered groupings of the test dataset.

Abstract: Methods, devices, and systems related to determining biometric data using an array of infrared (IR) illuminators are described. In an example, a method can include projecting a number of IR dots on a user using a dot projector and an array of IR illuminators, capturing an IR image of the number of IR dots using an IR camera, comparing a number of pixels of the captured IR image to a number of corresponding pixels of a baseline IR image using a processing resource, and determining biometric data of the user at least partially based on comparing the captured IR image to the baseline IR image using the processing resource.

Abstract: An intelligence-defined optical tunnel network system includes pods. Each pod includes optical add-drop sub-systems. Each optical add-drop sub-system includes a first transmission module and a second transmission module. The first transmission modules of the optical add-drop sub-systems are connected to each other for forming a first transmission ring. The second transmission modules of the optical add-drop sub-systems are connected to each other for forming a second transmission ring. Each first transmission module includes a multiplexer and an optical signal amplifier. The multiplexer is connected to a Top-of-Rack switch. The multiplexer is configured to receive, through input ports, upstream optical signals from the Top-of-Rack switch, and combine the upstream optical signals into a composite optical signal. The upstream optical signals have wavelengths respectively.

Abstract: The system includes a metric map creation unit configured to create a metric map using first image data received from a 3D sensor, an image processing unit configured to recognize an object by creating and classifying a point cloud using second image data received from an RGB camera; a probability-based map production unit configured to create an object location map and a spatial semantic map in a probabilistic expression method using a processing result of the image processing unit, a question creation unit configured to extract a portion of high uncertainty about an object class from a produced map on the basis of entropy and ask a user about the portion, and a map update unit configured to receive a response from the user and update a probability distribution for spatial information according to a change in probability distribution for classification of the object.

Abstract: A pulsed light communication device has a plurality of indicator light emitting diodes emitting diodes emitting at least one of a plurality of wavelengths of colored light to correspond to a designated color assigned to a security level for a network. A continuous uninterrupted modulated pulsed light emitting diode light signal may be generated having a sensitivity threshold detection level exceeding minimal parameters of a photodetector.

Abstract: Embodiments of the present disclosure provide systems and methods for training a machine-learning model for predicting emotions from received media data. Methods according to the present disclosure include displaying a user interface. The user interface includes a predefined media content, a plurality of predefined emotion tags, and a user interface control for controlling a recording of the user imitating the predefined media content. Methods can further include receiving, from a user, a selection of one or more emotion tags from the plurality of predefined emotion tags, receiving the recording of the user imitating the predefined media content, storing the recording in association with the selected one or more emotion tags, and training, based on the recording, the machine-learning model configured to receive input media data and predict an emotion based on the input media data.

Abstract: Aspects of the invention provide, in some aspects, a method of face recognition that includes receiving plural frames of a video stream imaging a candidate individual, e.g., in the field of view of a camera, and generating for each of those frames a score of the image and/or of the candidate therein. This can include, for example, a score (or count) indicative of the number of individuals present in the frame, a pose of the candidate individual (e.g., face-on or otherwise), blur in the image, and so forth. The method further includes selecting, based on the respective scores of the frames, a subset of the frames for matching by a face recognizer against a set of one or more images of designated individuals. That set may be of individuals approved for access, individuals to be prevented for access, or otherwise. An output is generated, according to the method, based on such matching by the face recognizer.

Abstract: A system for categorizing seeds of plants into hybrid and non-hybrid categories. Seeds sorted according to the disclosed system are also disclosed.

Abstract: Disclosed herein are methods and system for determining whether a user is wearing a mask, comprising receiving one or more infrared images depicting the user's face in one or more infrared spectral ranges and one or more visible light images depicting the user's face in visible light spectral range, registering the infrared image(s) to the visible light image(s), computing luminance values of a plurality of pixels relating to the user's face in the visible light image(s), computing infrared reflectiveness values of corresponding pixels in the registered infrared light image(s), computing, for each of the pixels, a difference between the luminance value and the infrared reflectiveness value and determining the user is genuine and not wearing a mask in case an aggregated difference aggregating the difference values of the pixels relating to the user's face exceeds a certain value.

Abstract: An embodiment proposed herein uses sparsification techniques to train the neural network with a high feature dimension that may yield desirable in-domain detection accuracy but may prune away dimensions in the output that are less important. Specifically, a sparsification vector is generated based on Gaussian distribution (or other probabilistic distribution) and is used to multiply with the higher dimension output to reduce the number of feature dimensions. The pruned output may be then used for the neural network to learn the sparsification vector. In this way, out-of-distribution detection accuracy can be improved.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: Systems, computer program products, and methods are described herein for character recognition in a digital image processing environment. The present invention is configured to electronically retrieve one or more documents from a document repository, wherein the one or more documents are in an image format; initiate one or more image super resolution algorithms on the one or more documents; generate, based on at least the one or more image super resolution algorithms, one or more high-resolution images associated with each of the one or more documents; initiate one or more image bottleneck ensembles (IBE) algorithms on the one or more high-resolution images; extract, using the one or more IBE algorithms, one or more features associated with the one or more high resolution images; and store the one or more features extracted from the one or more high resolution images in a feature repository.