Abstract: A product label printing and checking system, comprising data processing means configured to: control the printing of product labels; control the acquisition and reception of images of the printed labels by an image acquisition device such as an optical scanner; and check the printed labels for defects, wherein each printed label accords with a label format specification for that label, whereby the label has a common layout and comprises printed product-related information located in one or more regions on the label, wherein the data processing means includes a label checking module in which there is provided a reference image of the label, and the checking module is configured so that the acquired images are sequentially compared against the reference image according to pre-determined quality control indicators relating to the expected information content and location in the label regions, and wherein the label is flagged for review or rejection if it is non-compliant.

Abstract: A data analysis system includes: a memory; and a processor connected to the memory and that acquires data to be analyzed. The data to be analyzed includes parameters relating to a production element of a product produced in each of production batches; and an indicator for evaluating the product. The processor outputs, to the memory, the acquired data to be analyzed. The memory stores, in each of the production batches, the parameters and the indicator associated with the parameters. The processor calculates a correlation feature value for each of the production batches based on a correlation between the parameters and data of the correlation and the parameters. The processor causes a display to display the calculated feature value.

Abstract: There are provided methods and apparatus for in-loop artifact filtering. An apparatus includes an encoder for encoding an image region. The encoder has at least two filters for successively performing in-loop filtering to respectively reduce at least a first and a second type of quantization artifact.

Abstract: The invention relates to a method and an apparatus (10) for recognizing an article. This involves at least one shot (14), in particular in the form of a photograph, of the article being produced. The shot (14) is used to ascertain shot features (22) by means of a shot extraction algorithm (20). Stored article data (12) are used to ascertain article features (28) and to compare them with the shot features (22) in order to output association information (36). There is in particular provision according to the invention for a user rating (42) to be provided to improve both the shot extraction algorithm (20) and the article extraction algorithm (26). Alternatively or additionally, there is in particular provision according to the invention for both the shot extraction algorithm (20) and the article extraction algorithm (26) to be produced on the basis of interconnected, preferably weighted, data aggregation routines.

Abstract: The present disclosure relates to a system and method (10) for monitoring a set-up for manufacture and/or setting up for manufacture and/or tearing down after manufacture of a biopharmaceutical product. The method comprises: processing (S2, S3, S4) at least one image of a scene comprising the set-up for manufacture of the biopharmaceutical product. The processing of the at least one image comprises performing (S2) a first process on the at least one image for classifying first objects in the image, said first objects being devices such as clamps, pumps, valves and/or sensors and/or any other bio processing equipment. The first process comprising identifying, localizing and classifying the first objects in the image. A second process is performed (S3) on the at least one image for identifying and localizing connections in the images.

Abstract: An image identification method is provided, including: storing at least one normal state image of at least one test object; an automatic codec receiving the at least one normal state image to become a trained automatic codec; at least one camera device capturing at least one state image of the at least one test object; a computer device receiving the at least one state image, and the trained automatic codec performing feature extraction and reconstruction on the at least one state image to generate at least one reconstructed state image; and the computer device comparing the at least one state image and the at least one reconstructed state image, and determining whether the at least one state image is a normal state image. The present invention also provides an image identification system.

Abstract: Methods, systems, and computer program products are provided for cleaning noisy data from unlabeled datasets using autoencoders. A method includes receiving training data including noisy samples and other samples. An autoencoder network is trained based on the training data to increase a first metric based on the noisy samples and to reduce a second metric based on the other samples. Unlabeled data including unlabeled samples is received. A plurality of third outputs is generated by the autoencoder network based on the plurality of unlabeled samples. For each respective unlabeled sample, a respective third metric is determined based on the respective unlabeled sample and a respective third output, and whether to label the respective unlabeled sample as noisy or clean is determined based on the respective third metric and a threshold. Each respective unlabeled sample determined to be labeled as noisy is cleaned.

Abstract: The invention provides for a medical apparatus (100, 400, 600) comprising a memory (110) for storing machine executable instructions (120) and a processor (104) for controlling the medical apparatus. Execution of the machine executable instructions causes the processor to: receive (200) a medical image (122) descriptive of a three-dimensional anatomy of a subject (418); and provide (202) an image segmentation (124) by segmenting the medical image into multiple tissue regions (300, 302) using a model-based segmentation. The model-based segmentation assigns a tissue type to each of the multiple regions. The model-based segmentation has a surface mesh (304). The segmentation is corrected by using the tissue type assigned to each of the multiple regions to correct for partial volume effects at boundaries formed by the surface mesh between at least some of the multiple tissue regions.

Abstract: An action recognition method includes: obtaining original feature submaps of each of temporal frames on a plurality of convolutional channels by using a multi-channel convolutional layer; calculating, by using each of the temporal frames as a target temporal frame, motion information weights of the target temporal frame on the convolutional channels according to original feature submaps of the target temporal frame and original feature submaps of a next temporal frame, and obtaining motion information feature maps of the target temporal frame on the convolutional channels according to the motion information weights; performing temporal convolution on the motion information feature maps of the target temporal frame to obtain temporal motion feature maps of the target temporal frame; and recognizing an action type of a moving object in image data of the target temporal frame according to the temporal motion feature maps of the target temporal frame on the convolutional channels.

Abstract: An inspection system for determining wafer defects in semiconductor fabrication may include an image capturing device to capture a wafer image and a classification convolutional neural network (CNN) to determine a classification from a plurality of classes for the captured image. Each of the plurality of classes indicates a type of a defect in the wafer. The system may also include an encoder to encode to convert a training image into a feature vector; a cluster system to cluster the feature vector to generate soft labels for the training image; and a decoder to decode the feature vector into a re-generated image. The system may also include a classification system to determine a classification from the plurality of classes for the training image. The encoder and decoder may be formed from a CNN autoencoder. The classification CNN and the CNN autoencoder may each be a deep neural network.

Abstract: Disclosed is a system including a memory device having a plurality of physical memory segments and a processing device to perform operations that include, responsive to detecting a failure of a memory operation associated with a physical memory segment of the plurality of physical memory segments, quarantining the physical memory segment, responsive to quarantining the physical memory segment, performing one or more scanning operations on the physical memory segment, and determining, based on results of the one or more scanning operations, a viability status of the physical memory segment, wherein the viability status indicates an ability of the physical memory segment to store data.

Abstract: A method of controlling the quality of printed products by using a computer includes always producing multiple printed products with at least partially identical image content in a printing machine in a course of a print job to be completed. The multiple printed products are recorded by at least one image sensor and are sent to the computer as digital image data. The computer examines and assesses the digital image data in an image inspection process to find print defects and to sort out printed products that have been found to be unusable. The computer assesses the identical image content of the digital image data of at least two consecutive printed products and only takes into consideration such detected print defects that are present on the at least two assessed consecutive printed products.

Abstract: A device may receive, from a user device, connector panel information associated with a connector panel. The connector panel may provide a connection, via a port, for a service of a network. The device may receive an image that depicts a physical configuration of the connector panel. The device may process, using a port analysis model, the image to identify a port status of a port of the connector panel. The device may determine, based on the port status, that the port is available for the connection. The device may provide, to the user device, instructions for using the port for the connection. The device may obtain a verification that the connection has been established via the port. The device may perform one or more actions associated with providing the service.

Abstract: A method for small object detection in drone scene based on deep learning is provided, which includes: inputting images captured by a drone into a pre-trained generator based on an Unet network structure to output normal-light images; inputting the normal-light images into a object detection backbone network to output a plurality of multidimensional matrix feature maps, wherein the object detection backbone network integrates a channel attention mechanism and a spatial attention mechanism based on convolutional block Self-Block, and a 7*7 large convolutional kernel is used; inputting the plurality of multidimensional matrix feature maps into a BiFPN-S module of a feature pyramid for feature fusion, so as to output a plurality of corresponding feature maps for predicting objects of different sizes.

Abstract: An overlay metrology system may include a controller for receiving metrology data associated with a plurality of overlay targets on one or more samples; generating a reference metric for at least some of the plurality of overlay targets based on the metrology data, where the reference metric is associated with one or more properties of the respective overlay targets that contributes to overlay error; classifying the plurality of overlay targets into one or more groups based on the reference metrics calculated for the plurality of overlay targets; generating a reference image for at least some of the one or more groups; generating corrected metrology data using the associated reference image for at least some of the one or more groups; and generating overlay measurements for the plurality of overlay targets based on the corrected metrology data.

Abstract: Systems and methods are provided for generating a base visual score for each candidate image of a plurality of images received by a computing system, based on the scene type of each image. For each candidate image, the computing system multiplies the base visual score by a feature importance weight to generate a first visual score, adds respective scene type bonus points to the first visual score to generate a second visual score, and adds diversity scoring points to the second visual score to generate a final visual score for each candidate image. The computing system ranks the candidate images based on the final visual scores and provides a specified number of the top-ranked candidate images to be displayed on a display of the computing device.

Abstract: A model generation device (10) includes a specifier (103) that specifies partial moving images and a model generator (104) that generates an abnormality determination model. The partial moving images are included in a moving image acquired by imaging a production facility operable in multiple operation modes and are images for the respective operation modes. The model generator (104) generates, based on a time-series relationship between the partial moving images specified by the specifier (103) for the respective operation modes, the abnormality determination model for determination of whether an abnormality is present at the production facility based on the moving image acquired by imaging the production facility.

Abstract: A manufacturing method of a laminated sheet having a printed layer includes, given that a table that associates density data of a specific area included in image data with a burnup degree of the printed layer corresponding to the specific area at printing based on the image data is an image-burnup degree conversion table, identifying the image-burnup degree conversion table that satisfies a printing condition for the laminated sheet, acquiring printed image data that is image data for forming the printed layer of the laminated sheet, and calculating the burnup degree of the printed layer based on the image-burnup degree conversion table and the printed image data. Also disclosed is a laminated sheet having a printed layer, a system for manufacturing the laminated sheet, and a program for manufacturing the laminated sheet.

Abstract: A neural network is trained for use in a substrate thickness measurement system by obtaining ground truth thickness measurements of a top layer of a calibration substrate at a plurality of locations, each location at a defined position for a die being fabricated on the substrate. A plurality of color images of the calibration substrate are obtained, each color image corresponding to a region for a die being fabricated on the substrate. A neural network is trained to convert color images of die regions from an in-line substrate imager to thickness measurements for the top layer in the die region. The training is performed using training data that includes the plurality of color images and ground truth thickness measurements with each respective color image paired with a ground truth thickness measurement for the die region associated with the respective color image.

Abstract: According to the present invention, switching of the monitoring images matching the intention of the observer can be automatically performed for images from a plurality of image capturing apparatus, and the load about the job of the observer can be reduced. The image monitoring apparatus includes an estimating unit configured to estimate attention degrees of a user for a plurality of images acquired from the plurality of image capturing apparatuses, a designating unit configured to designate one of the acquired images as an image to be displayed in accordance with an instruction from the user, a learning unit configured to cause the estimating unit to learn so as to increase an attention degree of the designated image, and a selecting unit configured to select one of the plurality of images based on an attention degree of each estimated image.