A processor may receive an image and determine a number of foreground pixels in the image. The processor may obtain a result of optical character recognition (OCR) processing performed on the image. The processor may identify at least one bounding box surrounding at least one portion of text in the result and overlay the at least one bounding box on the image to form a masked image. The processor may determine a number of foreground pixels in the masked image and a decrease in the number of foreground pixels in the masked image relative to the number of foreground pixels in the image. Based on the decrease, the processor may modify an aspect of the OCR processing for subsequent image processing.

Disclosures herein teach applying a set of sections spanning a down-sampled version of an image of a road-scene to a low-fidelity classifier to determine a set of candidate sections for depicting one or more objects in a set of classes. The set of candidate sections of the down-sampled version may be mapped to a set of potential sectors in a high-fidelity version of the image. A high-fidelity classifier may be used to vet the set of potential sectors, determining the presence of one or more objects from the set of classes. The low-fidelity classifier may include a first Convolution Neural Network (CNN) trained on a first training set of down-sampled versions of cropped images of objects in the set of classes. Similarly, the high-fidelity classifier may include a second CNN trained on a second training set of high-fidelity versions of cropped images of objects in the set of classes.

A vision analytics and validation (VAV) system for providing an improved inspection of robotic assembly, the VAV system comprising a trained neural network three-way classifier, to classify each component as good, bad, or do not know, and an operator station configured to enable an operator to review an output of the trained neural network, and to determine whether a board including one or more “bad” or a “do not know” classified components passes review and is classified as good, or fails review and is classified as bad. In one embodiment, a retraining trigger to utilize the output of the operator station to train the trained neural network, based on the determination received from the operator station.

Methods, systems, and computer-readable storage media for receiving, by an automated review system, a legal document as a computer-readable file, and determining, by the automated review system, that the legal document is of a first type, and in response: converting the legal document to a set of images, extracting text data from one or more images in the set of images, the text data including sub-sets of text data, each sub-set of text data representing text in a respective clause of a set of clauses of the legal document, for each sub-set of text data receiving a prediction from a machine learning (ML) model in a set of ML models, the ML model being specific to a clause in the set of clauses, and outputting a set of predictions and respective prediction values for display in a user interface (UI).

Techniques are described for extracting key values from a document without having to rely on finding corresponding labels for the target keys within the extracted text of the document. Further the techniques do not rely on knowledge of the correlation between (a) the location of labels within a document, and (b) the location of the key values that correspond to the labels. Key values are extracted from a document by, identifying candidate values within the document, establishing “joint-candidate” sets from those candidate values, and using a trained machine learning mechanism to score each joint-candidate set of values. The highest scoring joint-candidate set is deemed to reflect the correct mapping of candidate values to target keys for the document.

A training request including an identifier that is indicative of a type of a machine learning (ML) model that is to be trained is received. A plurality of workers are maintained in a training pool, and a plurality of jobs are maintained in a queue of training jobs. Each worker is configured to train a particular type of ML model. Upon the training request being validated, a training job is created for the request and submitted to the queue of training jobs. For each type of ML model, a first metric and a second metric is obtained. A target metric is computed based on the first and the second metrics. The number of workers included in the training pool is modified based on the target metric.

Text profiles can be leveraged to select and configure models according to some examples described herein. In one example, a system can analyze a reference textual dataset and a target textual dataset using text-mining techniques to generate a first text profile and a second text profile, respectively. The first text profile can contain first metrics characterizing the reference textual dataset and the second text profile can contain second metrics characterizing the target textual dataset. The system can determine a similarity value by comparing the first text profile to the second text profile. The system can also receive a user selection of a model that is to be applied to the target textual dataset. The system can then generate an insight relating to an anticipated accuracy of the model on the target textual dataset based on the similarity value. The system can output the insight to the user.

The present disclosure provides systems, methods, and computer program products for generating a digital representation of a system from engineering documents of the system comprising one or more schematics and a components table. An example method can comprise (a) classifying, using a deep learning algorithm, (i) each of a plurality of symbols in the one or more schematics as a component and (ii) each group of related symbols as an assembly, (b) determining connections between the components and the assemblies, (c) associating a subset of the components and the assemblies with entries in the components table; and (d) generating the digital representation of the system from the components, the assemblies, the connections, and the associations. The digital representation of the system can comprise at least a digital model of the system and a machine-readable bill of materials.

A sensor device includes an array sensor having a plurality of detection elements arrayed in one or two dimensional manner, a signal processing unit configured to acquire a detection signal by the array sensor and perform signal processing, and a calculation unit. The calculation unit detects an object from the detection signal by the array sensor, and gives an instruction, to the signal processing unit, on region information generated on the basis of the detection of the object as region information regarding the acquisition of the detection signal from the array sensor or the signal processing for the detection signal.

Techniques for providing improved optical character recognition (OCR) for receipts are discussed herein. Some embodiments may provide for a system including one or more servers configured to perform receipt image cleanup, logo identification, and text extraction. The image cleanup may include transforming image data of the receipt by using image parameters values that optimize the logo identification, and performing logo identification using a comparison of the image data with training logos associated with merchants. When a merchant is identified, a second image clean up may be performed by using image parameter values optimized for text extraction. A receipt structure may be used to categorize the extracted text. Improved OCR accuracy is also achieved by applying on format rules of the receipt structure to the extracted text.