An NC-program conversion device includes an OCR processing unit that recognizes a character string from an input image; a first storage unit that stores one alphabetic letter and the number of digits of a number subsequent to the alphabetic letter, in an associated manner; a second storage unit that stores a program code that is composed of a combination of one alphabetic letter and a two-character number and an effective character that is composed of one alphabetic letter subsequent to the program code, in an associated manner; and a character-string segmenting unit that refers to the first storage unit and the second storage unit, to segment each line of the character string, which is recognized by the OCR processing unit, as program codes stored in the first storage unit in an associated manner.

A method and apparatus for detecting text are provided. The method includes: extracting features of a to-be-detected image; predicting using a character detection network a probability of each pixel point in the to-be-detected image being a character pixel point, and position information of each pixel point relative to a bounding box of a character including the pixel point when the pixel point is the character pixel point; determining position information of bounding boxes of candidate characters based on the prediction result of the character detection network; inputting the extracted features into a character map network, converting a feature map outputted by the character map network, and generating character vectors; determining a neighboring candidate character of each candidate character, and connecting each candidate character with an associated neighboring candidate character to form a character set; and determining a character area of the to-be-detected image.

A license plate detection and recognition system receives training data comprising images of license plates. The system prepares ground truth data from the training data based predefined parameters. The system trains a first machine learning algorithm based on the ground truth data to generate a license plate detection model. The license plate detection model is configured to detect one or more regions in the images. The one or more regions contains a candidate for a license plate. The LPDR system generates a bounding box for each region. The LPDR system trains a second machine learning algorithm based on the ground truth data and the license plate detection model to generate a license plate recognition model. The license plate recognition model generates a sequence of alphanumeric characters with a level of recognition confidence for the sequence.

In some implementations, a device may obtain an electronic document. The device may extract a portion of the electronic document. The device may obtain first text data associated with the portion. The device may determine whether the first text data corresponds to any of a plurality of text indicators of document types. The device may obtain, based on a determination that the first text data does not correspond to any of the plurality of text indicators, second text data associated with a greater portion of the electronic document that includes more than the first text data. The device may determine, using a machine learning model, a type of the electronic document based on the second text data. The device may determine whether the type of the electronic document differs from an expected document. The device may transmit a notification indicating that the electronic document is not the expected document.

A method and apparatus for training a character detector based on weak supervision, a character detection system and a computer readable storage medium are provided, wherein the method includes: inputting coarse-grained annotation information of a to-be-processed object, wherein the coarse-grained annotation information including a whole bounding outline of a word, text bar or line of the to-be-processed objected; dividing the whole bounding outline of the coarse-grained annotation information, to obtain a coarse bounding box of a character of the to-be-processed object; obtaining a predicted bounding box of the character of the to-be-processed object through a neural network model from the coarse-grained annotation information; and determining a fine bounding box of the character of the to-be-processed object as character-based annotation of the to-be-processed object, according to the coarse bounding box and the predicted bounding box.

An intelligent scoring method and system for a text objective question, the method comprising: acquiring an answer image of a text objective question (101); segmenting the answer image to obtain one or more segmentation results of an answer string to be identified (102); determining whether any of the segmentation results has the same number of characters as the standard answer (103); if no, the answer is determined to be wrong (106); otherwise, calculating identification confidence of the segmentation result having the same number of words as the standard answer, and/or calculating the identification confidence of respective characters in the segmentation result having the same number of words as the standard answer (104); determining whether the answer is correct according to the calculated identification confidence (105). The method can automatically score text objective questions, thus reducing consumption of human resource, and improving scoring efficiency and accuracy.

According to one embodiment, a character detection apparatus includes a feature extractor, a determiner and an integrator. The feature extractor extracts a feature value of an image including character strings. The determiner determines each priority of a plurality of different character detection schemes in accordance with character detection accuracy with respect to an image region having a feature corresponding to the feature value. The integrator integrates text line candidates of the character detection schemes, and selects, as a text line, one of the text line candidates detected by the character detection scheme with the highest priority if a superimposition degree indicating a ratio of a superimposed region among the text line candidates is no less than a first threshold value.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for neural network-based optical character recognition. An embodiment of the system may generate a set of bounding boxes based on reshaped image portions that correspond to image data of a source image. The system may merge any intersecting bounding boxes into a merged bounding box to generate a set of merged bounding boxes indicative of image data portions that likely portray one or more words. Each merged bounding box may be fed by the system into a neural network to identify one or more words of the source image represented in the respective merged bounding box. The one or more identified words may be displayed by the system according to a standardized font and a confidence score.

Systems for performing signature-based techniques on document during a document upload process are disclosed. During the document upload process, a document upload application may determine whether signature is required on the document to be uploaded and whether the required signature is present on the document. Upon determining that a signature is required but is missing from the document, document upload application may retrieve and merge the required signature with the document as part of the document upload process.

A picture processing method performed by an electronic device, includes: acquiring a target picture; performing region type detection and character recognition on the target picture by combining a target type detection technology and an optical character recognition technology, determining a target region to be occluded from the target picture based on a result of the region type detection and a result of the character recognition, and occluding the target region in the target picture.