A method for examining and counting incoming materials includes receiving three-dimensional scanned images of the incoming materials, wherein the three-dimensional scanned image is taken by an X-ray machine. The three-dimensional scanned image is preprocessed, each type of material is identified by a pre-trained material classification model and other information relevant thereto is collected, and a first total number of materials of each type is counted to obtain the total number of materials of each type.

A technology for cleaning a training data set for a neural network using dirty training data starts by accessing a labeled training data set that comprises relatively dirty labeled data elements. The labeled training data set is divided into a first subset A and a second subset B. The procedure includes cycling between the subsets A and B, including producing refined model-filtered subsets of subsets A and B to provide a cleaned data set. Each refined model-filtered subset can have improved cleanliness and increased numbers of elements.

The present invention describes a method for training a detector (16) of the presence of an anomaly in a signal relating to the environment acquired by a sensor (14) having several signal parts grouping together a set of successive temporal samples, the detector (16) comprising: a characteristic extraction module (21) for applying an extraction function to the signal, and a detection module (22) for detecting the presence of an anomaly in a signal part and for applying a detection function to a characteristic representing the signal part, to determine whether an anomaly is present, the training method comprising: obtaining the extraction function using self-supervised learning, and obtaining the detection function using semi-supervised learning.

A vehicle-mounted communication system includes multiple switch devices that are mounted on a vehicle. At least one of the multiple switch devices is capable of performing intermediate processing that is intermediate information processing on measurement information from a sensor that is mounted on the vehicle. The at least one of the multiple switch devices transmits at least information that is generated by the intermediate processing or the measurement information as relay information to another switch device. The other switch device transmits the relay information that is transmitted to a functional unit that is to perform a target process by using the relay information.

The present disclosure relates to systems, methods, and non-transitory computer-readable media for dynamically utilizing, in potentially real time, anomaly pattern detection to optimize operational processes relating to well construction or subterranean drilling. For example, the disclosed systems use time-series data combined with rig states to automatically detect and split similar operations. Subsequently, the disclosed systems identify operation anomalies from a field-data collection utilizing an automated anomaly detection workflow. The automated anomaly detection workflow can identify operation anomalies at a more granular level by determining which process behavior contributes to the operation anomaly (e.g., according to corresponding process probabilities for a given operation). In addition, the disclosed systems can present graphical representations of operation anomalies, process behaviors (procedural curves), and/or corresponding process probabilities in an intuitive, user-friendly manner.

A method for determining a threshold for spike detection in a noisy signal in real-time is provided. The method includes estimating a current variability of noise in the noisy signal according to a measured instantaneous value and a window of previous instantaneous values using a sigma-delta control loop, determining the threshold based on the estimated variability of the noise; and repeating the steps to update the estimate of the variability of the noise and adjust the threshold in real-time as the noisy signal changes. A non-transitory computer-readable storage medium for executing the method on a processing unit, and a low-power digital system implementing the method are also provided.

A system and method are provided to set an automatic window of interest based on learning data analysis. The system and method include a determination engine executed by a processor. The system and method include capturing catheter channel data from electrophysical studies and evaluating the catheter channel data using a machine learning tool to learn window of interest settings used within the electrophysical studies. The system and method further include automatically setting a window of interest for a mapping procedure based on the window of interest settings.

Method for acquiring data for detecting damage to a bearing, comprising the following steps: obtaining a vibration signal from the bearing over a period of time, determining spectrograms of said vibration signal at various instants of the period of time, detecting the peaks on each spectrogram, from among the detected peaks, retaining the peaks which, in spectrograms corresponding to successive instants and transformed through synchronous resampling at a rotational speed of the bearing, would be present at the same frequency, with a predetermined tolerance, identifying spectral lines corresponding to the retained peaks, and storing information representative of said spectral lines in a record able to be used by a unit for detecting damage to the bearing.

A cycling detection method, an electronic device, and a computer-readable storage medium are provided. The cycling detection method includes: obtaining an acceleration signal and an angular velocity signal collected by a wearable device of a user's foot, and analyzing a time-domain feature of the acceleration signal, a frequency-domain feature of the acceleration signal, a time-domain feature of the angular velocity signal, and/or a frequency-domain feature of the angular velocity signal; filtering the acceleration signal and the angular velocity signal to obtain a target acceleration signal and a target angular velocity signal of the user's cycling if it is determined, based on an analysis result, that a current status of the user is a cycling state; and determining a behavior feature of the user's cycling based on the target acceleration signal and the target angular velocity signal, where the behavior feature includes a cadence feature and a foot posture feature.

An apparatus and method are providing for generating a hash in video in which a sample series of temporal difference are sampled in an image order. A temporal averaging is performed and a rate of change is detected to identify as distinctive events regions of high rate of change. Images having a distinctive event are labelled as distinctive images. For each image, the temporal spacing in images is calculated between that image and other distinctive images to provide a set of temporal spacings for that image; and a hash is derived for that image from that set of temporal spacings.