Various embodiments of the teachings herein include a method for training and updating a backend-side classifier comprising: receiving, in a backend-device, from at least one vehicle, classification data along with a respective classification result generated by a vehicle-side classifier; and training the backend-side classifier using the classification data and, if available, a corrected respective classification result as annotation.

Systems, methods, and other embodiments associated with detecting severity levels of risk in an electronic correspondence are described. In one embodiment, a method includes inputting, into a memory, a target electronic correspondence that has been classified as being litigious by a machine learning classifier. An artificial intelligence rule-based technique is applied to the target electronic correspondence that identifies high and medium risk level keywords. The technique is also configured to generate a litigious score based on a sum of term frequencies-inverse document frequencies using the remaining keywords. An electronic notice is transmitted to a remote computer over a communication network that identifies the target electronic correspondence and the level of litigation risk.

Various embodiments of the teachings herein include a method for training and updating a backend-side classifier comprising: receiving, in a backend-device, from at least one vehicle, classification data along with a respective classification result generated by a vehicle-side classifier; and training the backend-side classifier using the classification data and, if available, a corrected respective classification result as annotation.

Mechanisms are provided to implement a hardened ensemble artificial intelligence (AI) model generator. The hardened ensemble AI model generator co-trains at least two AI models. The hardened ensemble AI model generator modifies, based on a comparison of the at least two AI models, a loss surface of one or more of the at least two AI models to prevent an adversarial attack on one AI model, in the at least two AI models, transferring to another AI model in the at least two AI models, to thereby generate one or more modified AI models. At least one of the one or more modified AI models then processes an input to generate an output result.

A defect detection device includes an image capturing component for capturing one or more images of an object to be inspected; a motion component configured to grasp or manipulate the object or the image capturing component; and a computing device configured to perform a defect detection method, including determining a plurality of first image capturing poses of the object to be inspected; determining a first defect probability for each particular first image capturing pose; establishing a probability matrix based on the first defect probabilities; subdividing the probability matrix into a plurality of submatrices according to preset dimensions for each submatrix; determining a second defect probability for each of the plurality of submatrices; setting a maximum value of the second defect probabilities as a third defect probability of the object to be inspected; and comparing the third defect probability to a threshold to determine whether the object is defective.

A document processing framework (DPF) for robotic process automation (RPA) is provided. The DPF may allow plug-and-play use of different vendor products on same platform, where users can setup a basic schema for document processing and document understanding workflow. The DPF may allow users to define a taxonomy, digitize a file, classify the file into one or more document types, validate the classification, extract data, validate the extracted data, train classifiers, and/or train extractors. A public package may be provided that can be used by software developers to manage the DPF and build their own classifier and extractor components.

Disclosed is an object-detection system configured to utilize multiple object-detection models that generate respective sets of object-detection conclusions to detect objects of interest within images of scenes. The object-detection system is configured to implement a series of functions to reconcile any discrepancies that exist in its multiple sets of object-detection conclusions in order to generate one set of conclusions for each perceived object of interest within a given image.

In a method of determining the position of an object, raw image data of the sky is recorded using a celestial imaging unit. The last known position, orientation, date, and time data of the object are obtained, and the position of a celestial body is measured. A latitude and longitude of the object is determined by matching the measured celestial body position to the expected celestial body position based on the input parameters. A system for determining a new position of an object comprises a celestial imaging unit configured to record image data of the sky, a signal processing unit, and a signal processing unit configured to receive and store in memory the image data received from the celestial imaging unit. The signal processing unit filters the image to find the positions of celestial objects in the sky. The signal processing unit is further configured to use roll and pitch from an IMU, and date and time from a clock to determine the object's position (latitude and longitude).

Systems, methods, and other embodiments associated with a machine learning system that monitors and detects risk in electronic correspondence related to a construction project are described. In one embodiment, a method includes monitoring email communications over a network to identify an email; tokenizing text from the email into a plurality of words and initiating a machine learning classifier configured to identify construction terminology and to classify text with a risk as being litigious or non-litigious. The machine learning classifier processes the words from the email by at least corresponding the words to a set of defined litigious vocabulary and defined non-litigious vocabulary. The email is labeled as litigious or non-litigious. An electronic notice is generated and transmitted to a remote device in response to the email being labeled as being litigious to provide an alert in near-real time in relation to receiving the email over the network.

Disclosed herein systems, processors, or computer-readable media configured with instructions to: receive transmission and/or reflection images of a tissue of a subject, wherein the images are generated from acoustic signals derived from acoustic waveforms transmitted through the tissue; provide a set of prognostic parameters associated with a user selected region of interest; wherein the set of prognostic parameters comprises sound propagation metrics characterizing sound propagation within a tissue; wherein the set of prognostic parameters corresponds to inputs into a tissue classifier model; wherein the set of prognostic parameters comprises a plurality of subsets of related feature groupings; and determine a type of tissue of the subject based on said plurality of subsets of related feature groupings using the classifier model, wherein the type of tissue is a cancerous tumor, a fibroadenoma, a cyst, a nonspecific benign mass, and an unidentifiable mass.