CN114972880A - Label identification method and device, electronic equipment and storage medium - Google Patents
Label identification method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114972880A CN114972880A CN202210676549.7A CN202210676549A CN114972880A CN 114972880 A CN114972880 A CN 114972880A CN 202210676549 A CN202210676549 A CN 202210676549A CN 114972880 A CN114972880 A CN 114972880A
- Authority
- CN
- China
- Prior art keywords
- image
- label
- predicted
- labels
- recognized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a label identification method, a label identification device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified; predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels; a target label of the item is determined based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model. Namely, the embodiment of the invention improves the quality of the labeling image by preprocessing the labeling image, further improves the accuracy of label identification, performs label identification on the image to be identified through a label prediction network to obtain a plurality of predicted labels, and screens or determines the plurality of predicted labels by combining a label determination model to obtain a target label, thereby improving the efficiency and accuracy of identification, and avoiding the influence of the environment on the identification effect under the condition of no personnel participation.
Description
Technical Field
The present invention relates to computer technologies, and in particular, to a tag identification method and apparatus, an electronic device, and a storage medium.
Background
The quality inspection link in the production line of a factory is generally finished by quality inspectors to prevent poor products from flowing into the market, and the quality inspection link relates to the procedures of accessory labeling type confirmation, product part assembly quality confirmation, product surface scratch defect detection and the like until a plurality of quality inspectors are matched and finished. The mode of producing in mixture is adopted mostly on the production line of mill, and some products are only that inside spare part model is different, and it is not very big from making the external dimension difference, lead to the product model, paste a mark kind numerous, and such as energy consumption post, product data plate, warning subsides, power consumption etc. the font is little, the serial number is long, the quality testing personnel check consuming time, the different subsides of different products mark very easily obscures, the condition of lou examining, false retrieval also sometimes takes place, trace back also more difficult. In order to solve the problems, in the prior art, an Optical Character Recognition (OCR) method is mostly adopted to replace manual work for automatically recognizing the text content in the labeling image, so that the detection efficiency and the accuracy can be effectively improved.
Disclosure of Invention
The invention provides a tag identification method, a tag identification device, electronic equipment and a storage medium, which are used for automatically identifying contents in a tag under the condition of no personnel participation.
In a first aspect, the embodiment of the invention provides a method for acquiring a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
determining, with a tag determination model, a target tag for the item based on the plurality of predicted tags and the confidence level of each predicted tag.
Further, the predicting a label in the image to be recognized by using the label prediction network to obtain a plurality of predicted labels in the image to be recognized and a confidence level of each predicted label in the plurality of predicted labels includes:
extracting a characteristic vector of a detection area in the image to be identified by utilizing the convolutional network;
and extracting character features of the feature vector according to the semantic relation of the feature vector by using the long and short memory network, and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
Further, the extracting the feature vector of the detection region in the image to be identified by using the convolutional network includes:
inputting the image to be identified into a convolutional layer to extract image features in the image to be identified, and using the deconvolution layer to perform upsampling on the convolutional layer to obtain a convolutional image, wherein a correlation exists between the convolutional image and pixels of the image to be identified;
and carrying out pixel classification on the convolution image to obtain a two-dimensional thermodynamic diagram of the image to be identified, and carrying out image segmentation on the image to be identified according to the two-dimensional thermodynamic diagram to obtain a feature vector of the detection area.
Further, the extracting, by using the long and short memory network, the text features of the feature vector according to the semantic relationship of the feature vector, and predicting the label in the image to be recognized according to the text features to obtain a plurality of predicted labels in the image to be recognized and a confidence of each predicted label in the plurality of predicted labels includes:
inputting the feature vector of the detection area into the long and short memory network to obtain the semantic relation of the feature vector, and extracting the character features of the semantic relation;
and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
Further, after determining the target label of the article based on the plurality of predicted labels and the confidence of each predicted label by using the label determination model, the method further includes:
extracting attribute information from the target label of the article, and storing the attribute information and the labeling image in an information database in a correlation manner, so that the attribute information of the article is inquired according to the labeling image.
Further, the tag determination model is obtained as follows:
obtaining a training sample, wherein the training sample comprises a plurality of identification images, and each identification image comprises a plurality of prediction labels and confidence degrees of each prediction label in the plurality of prediction labels;
screening a plurality of prediction labels in each recognition image in the training sample and the confidence coefficient of each prediction label in the plurality of prediction labels by using a time sequence class classification model to obtain a target label corresponding to each recognition image;
calculating a loss function according to the target label corresponding to each identification image and the actual label corresponding to each identification image;
and performing back propagation according to the loss function to optimize the time sequence class classification model to obtain the label determination model.
Further, after predicting the label in the image to be recognized by using a label prediction network, the method further includes:
and inputting the image to be identified into the label prediction network to obtain identification error information of the image to be identified, and performing fault early warning according to the identification error information.
In a second aspect, an embodiment of the present invention further provides a tag identification apparatus, where the apparatus includes:
the image acquisition module is used for acquiring a labeling image of an article and preprocessing the labeling image to obtain an image to be identified;
the network prediction module is used for predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
a tag determination module to determine a target tag of the item based on the plurality of predicted tags and the confidence level of each predicted tag using a tag determination model.
In a third aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the tag identification method.
In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the tag identification method.
In the embodiment of the invention, the image to be identified is obtained by obtaining the labeling image of the article and preprocessing the labeling image; predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels; a target label of the item is determined based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model. Namely, the embodiment of the invention improves the quality of the labeling image by preprocessing the labeling image, further improves the accuracy of label identification, performs label identification on the image to be identified through a label prediction network to obtain a plurality of predicted labels, and screens or determines the plurality of predicted labels by combining a label determination model to obtain a target label, thereby improving the efficiency and accuracy of identification, and avoiding the influence of the environment on the identification effect under the condition of no personnel participation.
Drawings
Fig. 1 is a schematic flow chart of a tag identification method according to an embodiment of the present invention;
fig. 2 is another schematic flow chart of a tag identification method according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a tag identification device according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Fig. 1 is a schematic flowchart of a tag identification method according to an embodiment of the present invention, where the method may be executed by a tag identification apparatus according to an embodiment of the present invention, and the apparatus may be implemented in software and/or hardware. In a specific embodiment, the apparatus may be integrated in an electronic device, which may be, for example, a server. The following embodiments will be described by taking as an example that the apparatus is integrated in an electronic device, and referring to fig. 1, the method may specifically include the following steps:
s110, obtaining a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
for example, the labeling image may be from an image acquisition device, the image acquisition device may be a camera, a video recorder, or other devices with an image acquisition function, and the labeling image may be an image acquired in real time or an image acquired in advance; when the labeling image is an image acquired in real time, the labeling image can be an image acquired by image acquisition equipment in real time on a factory production line or a preset area of an article, and in this case, the label of the article in the current scene is identified in real time by using the real-time image; when the labeling image is a pre-collected image, the labeling image can be any image in the labeling image library, in this case, the pre-collected labeling image is used for identifying the article label in the current scene, and the condition that the content of the article label in the pre-collected labeling image cannot be searched or lost can be realized. The image to be identified can be an image obtained by preprocessing the labeling image by utilizing a preprocessing technology, and has higher definition than the labeling image.
In the specific implementation, after the labeling image of the article is obtained, the labeling image can be preprocessed to improve the quality of the labeling image and obtain an image to be identified; such as: the obtained labeling image can be subjected to image processing methods such as graying, image enhancement, inclination monitoring and correction, Gaussian filtering and the like, image noise is eliminated, computing resources are saved, image quality and image processing speed are improved, and label detection and identification accuracy and precision of an article are guaranteed. The method comprises the steps of scanning a bar code of an article to obtain a serial number of the article, wherein the serial number can be a serial number of a position where the article is placed or stored, or a serial number corresponding to a produced article, simultaneously obtaining a labeling image of the article by using image acquisition equipment, wherein the labeling image comprises image content of a label to be identified of the article, and storing the serial number of the article and the labeling image of the article in a correlation manner to determine the article corresponding to the identified label.
S120, predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
for example, the tag prediction network may be a neural network for identifying a tag of an article in an image to be identified, where the tag prediction network includes a convolutional network and a long-short memory network, where the convolutional network is configured to extract image features in the image to be identified, and perform image segmentation at a pixel level according to the image features in the image to be identified, so as to obtain a feature vector in a detection region. The detection area may be an image area corresponding to a tag in an image to be identified, and the feature vector in the detection area may be a character feature sequence vector extracted by image features in the detection area. The long and short memory network can be used for capturing forward information and backward information of feature vectors in a detection area output by the convolutional network, fusing the feature vectors of the detection area on the basis of the convolutional layer to extract context features of a character sequence, and predicting label distribution of each feature vector in the feature sequence and probability distribution of each row of features. The multiple prediction labels in the image to be recognized may be multiple prediction results of labels in an image region corresponding to the labels in the image to be recognized output by the long and short memory network, and the confidence of each prediction label in the multiple prediction labels may be a probability value corresponding to the multiple prediction results of the labels in the image region corresponding to the labels in the image to be recognized output by the long and short memory network.
In the concrete implementation, the image to be recognized is input into a convolution network in a label prediction network, image characteristic information in the image to be recognized is extracted from the convolution network, the convolution layer in the convolution network is up-sampled to obtain an image with the size consistent with that of the image to be recognized, the image to be recognized and the up-sampled image are in one-to-one correspondence, the up-sampled image is classified at a pixel level, the type of each pixel is recovered from abstract characteristics, and a two-dimensional thermodynamic diagram after the classification at the pixel level is obtained. And performing target detection and edge detection based on the two-dimensional thermodynamic diagram, performing pixel-level image segmentation on the image according to the label features in the thermodynamic diagram, realizing accurate positioning and accurate segmentation of a detection area in the image to be recognized, acquiring image features in the detection area, performing text detection according to the image features in the monitoring area, extracting character feature sequence vectors, and obtaining feature vectors of the image to be recognized. Capturing forward information and backward information of a text sequence from a feature vector of an image to be recognized by using a long and short memory network, extracting context features of the text sequence based on the feature vector of the image to be recognized, and predicting a label text in a detection area by using upper and lower features of the text sequence to obtain a plurality of predicted labels in the image to be recognized and a confidence coefficient of each predicted label in the plurality of predicted labels.
In the embodiment of the invention, a plurality of convolution layers can be used in the convolution network to extract the image characteristics in the image to be identified, the convolution network can adopt a full convolution network FCN to realize the convolution of the image with any size and obtain the image marked with the probability value of each pixel category, and the characteristic diagram of the last convolution layer in the convolution network is subjected to up-sampling, so that the image passing through the convolution layers can recover the size same as the size of the image to be identified, and the incidence relation between the pixels is reserved for two images with the same size, wherein each pixel can generate a prediction.
And S130, determining a target label of the article based on the plurality of predicted labels and the confidence coefficient of each predicted label by using a label determination model.
In a specific implementation, the plurality of predicted tags and the confidence level of each predicted tag may be input into a tag determination model, the output of the tag determination model may be content in a target tag, the target tag of an article may be obtained according to the confidence level of each predicted tag, the target tag may be text information in a detection region in an image to be recognized, and the text information may be key information of the article, such as: attribute information of the article, function information of the article, and parameters of the article. The label determination model may be trained from training samples. The method can also be used for determining the label content in the target label of the article through blank removal and duplication removal operations, so that the quality of the target label of the article can be improved.
In the embodiment of the invention, the image to be identified is obtained by obtaining the labeling image of the article and preprocessing the labeling image; predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels; a target label of the item is determined based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model. Namely, the embodiment of the invention improves the quality of the labeling image by preprocessing the labeling image, further improves the accuracy of label identification, performs label identification on the image to be identified through a label prediction network to obtain a plurality of predicted labels, and screens or determines the plurality of predicted labels by combining a label determination model to obtain a target label, thereby improving the efficiency and accuracy of identification, and avoiding the influence of the environment on the identification effect under the condition of no personnel participation.
The tag identification method provided in the embodiment of the present invention is further described below, and as shown in fig. 2, the method may specifically include the following steps:
s210, obtaining a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
s220, extracting a feature vector of a detection area in the image to be identified by using a convolutional network;
in specific implementation, the detection region in the image to be recognized may be a label region in the image to be recognized, which is determined by image segmentation according to different pixel levels after extracting the features of the convolutional network. The feature vector of the detection area in the image to be recognized may be a vector formed by extracting a text feature sequence in the detection area after text detection is performed according to the image features of the detection area in the image to be recognized. Inputting the image to be recognized into a convolution network to obtain the image characteristics of the image to be recognized, performing up-sampling on the image after the characteristic extraction based on the image characteristics of the image to be recognized, enabling the size of the image after the characteristic extraction to be consistent with that of the image to be recognized, and keeping the relationship between the image after the characteristic extraction and the pixels of the image to be recognized. And carrying out pixel classification on the image after the characteristic extraction to obtain a two-dimensional thermodynamic diagram of the image to be identified. And performing pixel-level image segmentation on the image to be recognized according to the two-dimensional thermodynamic diagram, determining a detection area in the image to be recognized according to the label characteristics, performing text detection on the image characteristics in the detection area, and extracting the characteristic vector of the detection area.
Further, the convolutional network comprises a convolutional layer and a deconvolution layer, and the extracting of the feature vector of the detection region in the image to be identified by using the convolutional network comprises:
inputting an image to be identified into a convolutional layer to extract image characteristics in the image to be identified, and performing up-sampling on the convolutional layer by using an anti-convolutional layer to obtain a convolutional image, wherein correlation exists between the convolutional image and pixels of the image to be identified;
and carrying out pixel classification on the convolution image to obtain a two-dimensional thermodynamic diagram of the image to be recognized, and carrying out image segmentation on the image to be recognized according to the two-dimensional thermodynamic diagram to obtain a feature vector of the detection area.
For example, the image feature in the image to be recognized may be image feature information extracted from the image to be recognized by using the convolutional layer, and may be boundary information, gray scale information, character information, and the like of an object in the image to be recognized. The convolutional network comprises a convolutional layer and a deconvolution layer, wherein the convolutional layer is used for extracting image features in the image to be identified, and the deconvolution layer is used for carrying out up-sampling on the convolutional layer. The convolution image may be an image that is output after upsampling by an deconvolution layer and that maintains the same size as the image to be recognized, the convolution image preserving the relationship with the image pixels to be recognized. The two-dimensional thermodynamic diagrams can be used for distinguishing images of different pixel categories in the image to be recognized, the area positions of different pixels can be visually seen, and the detection area can be accurately positioned and segmented by the aid of the two-dimensional thermodynamic diagrams.
In specific implementation, the image to be recognized is input into a convolution network to obtain the image characteristics of the image to be recognized, the image after the characteristic extraction is subjected to up-sampling based on the image characteristics of the image to be recognized, so that the size of the image after the characteristic extraction is consistent with that of the image to be recognized, and the relationship between the image after the characteristic extraction and the pixels of the image to be recognized is reserved. And carrying out pixel classification on the image after the characteristic extraction to obtain a two-dimensional thermodynamic diagram of the image to be identified. And performing pixel-level image segmentation on the image to be recognized according to the two-dimensional thermodynamic diagram, determining a detection area in the image to be recognized according to the label characteristics, performing text detection on the image characteristics in the detection area, and extracting character characteristics in the image characteristics in the detection area to obtain a characteristic vector of the detection area.
S230, extracting character features of the feature vectors according to the semantic relation of the feature vectors by using a long and short memory network, and predicting the labels in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
for example, the semantic relationship of the feature vector may be forward information and backward information of the feature vector in the detection region, where the forward information may be forward semantic transfer information corresponding to the content of the tag to be predicted in the detection region, that is, forward semantic information associated with the content of the tag to be predicted in the detection region, and the backward information may be backward semantic transfer information corresponding to the content of the tag to be predicted in the detection region, that is, forward semantic information associated with the content of the tag to be predicted in the detection region, such as: the tag information in the image is 123456X890, where "X" is the tag content to be predicted, "12345" is the forward information, and "890" is the backward information, where the forward information may also be the feature information corresponding to the forward content of the tag content to be predicted, and the backward information may also be the feature information corresponding to the backward content of the tag content to be predicted. The text feature of the feature vector may be a feature vector text sequence arrangement rule.
In the specific implementation, the feature vector of the detection region is input into a long and short memory network, the long and short memory network captures forward information and backward information of a character feature sequence in the feature vector of the detection region output by a convolutional network to obtain a semantic relation of the feature vector, and character features of the content of the label to be predicted in the detection region are extracted after feature fusion is carried out according to the semantic relation of the feature vector and the feature vector of the detection region. And predicting the label in the image to be recognized according to the character characteristics to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels. The long and short memory networks are neural networks trained through training samples, and the content of the labels to be predicted in the detection area can be predicted through character features. The content of the to-be-predicted label can be the label content in a detection area in the to-be-recognized image which is not recognized through a convolutional network, and content prediction is required to be performed according to a length memory network, wherein the content of the to-be-predicted label can be a character or a plurality of characters, when the content of the to-be-predicted label is a character, the position of the character can be predicted, namely the position of the character can be predicted through a plurality of prediction labels and the confidence coefficient of each prediction label in the plurality of prediction labels; when the content of the label to be predicted is a plurality of characters, the character positions can be respectively predicted, then the predicted characters and the confidence degrees of the character positions are respectively obtained, and the confidence degrees of the plurality of predicted labels and each predicted label in the plurality of predicted labels are obtained by calculating according to the predicted characters and the confidence degrees of the character positions.
Further, extracting character features of the feature vector according to the semantic relation of the feature vector by using a long and short memory network, predicting a label in the image to be recognized according to the character features, and obtaining a plurality of predicted labels in the image to be recognized and a confidence coefficient of each predicted label in the plurality of predicted labels, wherein the method comprises the following steps:
inputting the feature vectors of the detection area into a long and short memory network to obtain the semantic relation of the feature vectors, and extracting character features of the semantic relation;
and predicting the labels in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
In the specific implementation, the feature vector of the detection region is input into a long and short memory network, the long and short memory network captures forward information and backward information of a character feature sequence in the feature vector of the detection region output by a convolutional network to obtain a semantic relation of the feature vector, and character features of the content of the label to be predicted in the detection region are extracted after feature fusion is carried out according to the semantic relation of the feature vector and the feature vector of the detection region. And predicting the label in the image to be recognized according to the character characteristics to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels. The long and short memory networks are neural networks trained through training samples, and the content of the labels to be predicted in the detection area can be predicted through character features.
Further, after predicting the label in the image to be recognized by using the label prediction network, the method further includes:
and inputting the image to be identified into a label prediction network to obtain identification error information of the image to be identified, and performing fault early warning according to the identification error information.
For example, the identification error information may be information correspondingly sent when the label content in the image to be identified is not predicted by the label prediction network, and the identification error information is used to prompt that the label in the image to be identified cannot be identified or that the label of the image to be identified cannot meet the prediction condition of the label prediction network, and trigger an alarm function to perform fault early warning according to the identification error information, and the label prediction network needs to be retrained or the image to be identified needs to be re-photographed or preprocessed.
S240, determining a target label of the article based on the plurality of predicted labels and the confidence coefficient of each predicted label by using a label determination model;
further, the tag determination model is obtained as follows:
obtaining a training sample, wherein the training sample comprises a plurality of identification images, and each identification image comprises a plurality of prediction labels and a confidence coefficient of each prediction label in the plurality of prediction labels;
screening a plurality of prediction labels in each recognition image in a training sample and the confidence coefficient of each prediction label in the plurality of prediction labels by using a time sequence class classification model to obtain a target label corresponding to each recognition image;
calculating a loss function according to the target label corresponding to each identification image and the actual label corresponding to each identification image;
and performing back propagation according to the loss function to optimize the time sequence class classification model to obtain a label determination model.
For example, the training sample may be an image set which is collected according to the label identification content and contains the label identification content, each identification image which is predicted by a plurality of identification images in the training sample through a label prediction network contains a plurality of prediction labels and the confidence of each prediction label in the plurality of prediction labels is labeled in advance, and an actual label corresponding to each identification image is labeled in each identification image. The time sequence class classification model can be a neural network according to a time sequence class, and is used for screening or identifying the plurality of prediction tags and the confidence coefficient of each prediction tag in the plurality of prediction tags to obtain a target tag corresponding to the span-identified image. The actual label corresponding to each recognition image may be the actual label content corresponding to each recognition image in the training sample, and is used for performing model training on the label determination model.
In the specific implementation, a plurality of identification images in a training sample are obtained, the identification images are preprocessed in advance, the preprocessed identification images are input into a label prediction network, and a plurality of prediction labels of each identification image and the confidence coefficient of each prediction label in the prediction labels are obtained. And inputting the plurality of prediction labels of each recognition image and the confidence coefficient of each prediction label in the plurality of prediction labels into a time sequence class classification model for label determination to obtain a target label corresponding to each recognition image. And calculating a loss function according to the target label corresponding to each identification image and the actual label corresponding to each identification image to obtain the entropy value of the loss function. And determining whether the time sequence class classification model is converged according to the entropy of the loss function, performing back propagation according to the entropy of the mathematical function to optimize parameters of the time sequence class classification model until the entropy of the loss function is smaller than a preset entropy threshold, and determining that the time sequence class classification model is converged to obtain a label determination model.
And S250, extracting the attribute information from the target label of the article, and storing the attribute information and the labeling image into an information database in a correlation manner, so that the attribute information of the article is inquired according to the labeling image.
In a specific implementation, the attribute information may be information having an item attribute in target tag content of an item, such as: energy consumption model, nameplate number, comprehensive power consumption and the like. The information database can be a database for storing article related information, and can be used in a factory manufacturing execution system, and the bar code, the labeling image and the attribute information of the article are stored in the value information database in a correlated storage mode, so that the article can be traced and inquired later. After the target labels of the articles are subjected to label identification, the identified target labels can be processed by regular-case filtering, a search algorithm and the like, and attribute information of the articles in the target labels of the articles can be extracted.
In the embodiment of the invention, the image to be identified is obtained by acquiring the labeling image of the article and preprocessing the labeling image; predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels; a target label of the item is determined based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model. Namely, the embodiment of the invention improves the quality of the labeling image by preprocessing the labeling image, further improves the accuracy of label identification, performs label identification on the image to be identified through a label prediction network to obtain a plurality of predicted labels, and screens or determines the plurality of predicted labels by combining a label determination model to obtain a target label, thereby improving the efficiency and accuracy of identification, and avoiding the influence of the environment on the identification effect under the condition of no personnel participation.
Fig. 3 is a schematic structural diagram of a tag identification apparatus according to an embodiment of the present invention, and as shown in fig. 3, the tag identification apparatus includes:
the image acquisition module 310 is used for acquiring a labeling image of an article and preprocessing the labeling image to obtain an image to be identified;
a network prediction module 320, configured to predict, by using a tag prediction network, a tag in the image to be recognized, so as to obtain multiple predicted tags in the image to be recognized and a confidence of each of the multiple predicted tags;
a label determination module 330, configured to determine a target label of the item based on the plurality of predicted labels and the confidence of each predicted label using a label determination model.
In an embodiment, the predicting module 320 of the network includes a convolutional network and a long and short memory network, and the predicting module of the network predicts the tags in the image to be recognized by using the tag prediction network to obtain the multiple prediction tags in the image to be recognized and the confidence of each prediction tag in the multiple prediction tags, including:
extracting a characteristic vector of a detection area in the image to be identified by utilizing the convolutional network;
and extracting character features of the feature vector according to the semantic relation of the feature vector by using the long and short memory network, and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
In an embodiment, the network prediction module 320 includes a convolution layer and a deconvolution layer, and the extracting feature vectors of the detection regions in the image to be identified by using the convolution network includes:
inputting the image to be identified into a convolutional layer to extract image features in the image to be identified, and using the deconvolution layer to perform upsampling on the convolutional layer to obtain a convolutional image, wherein a correlation exists between the convolutional image and pixels of the image to be identified;
and carrying out pixel classification on the convolution image to obtain a two-dimensional thermodynamic diagram of the image to be identified, and carrying out image segmentation on the image to be identified according to the two-dimensional thermodynamic diagram to obtain a feature vector of the detection area.
In an embodiment, the extracting, by the network prediction module 320, the text features of the feature vector according to the semantic relationship of the feature vector by using the long and short memory network, and predicting the labels in the image to be recognized according to the text features to obtain the multiple prediction labels in the image to be recognized and the confidence of each prediction label in the multiple prediction labels includes:
inputting the feature vector of the detection area into the long and short memory network to obtain the semantic relation of the feature vector, and extracting the character features of the semantic relation;
and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
In one embodiment, after the tag determination module 330 determines the target tag of the article based on the plurality of predicted tags and the confidence of each predicted tag by using the tag determination model, the method further includes:
extracting attribute information from the target label of the article, and storing the attribute information and the labeling image in an information database in a correlation manner, so that the attribute information of the article is inquired according to the labeling image.
In an embodiment, the tag determination model in the tag determination module 330 is obtained as follows:
obtaining a training sample, wherein the training sample comprises a plurality of identification images, and each identification image comprises a plurality of prediction labels and confidence degrees of each prediction label in the plurality of prediction labels;
screening a plurality of prediction labels in each recognition image in the training sample and the confidence coefficient of each prediction label in the plurality of prediction labels by using a time sequence class classification model to obtain a target label corresponding to each recognition image;
calculating a loss function according to the target label corresponding to each identification image and the actual label corresponding to each identification image;
and performing back propagation according to the loss function to optimize the time sequence class classification model to obtain the label determination model.
In an embodiment, after the network prediction module 320 predicts the tag in the image to be recognized by using the tag prediction network, the method further includes:
and inputting the image to be recognized into the label prediction network to obtain recognition error information of the image to be recognized, and performing fault early warning according to the recognition error information.
According to the device provided by the embodiment of the invention, the image to be identified is obtained by acquiring the labeling image of the article and preprocessing the labeling image; predicting the labels in the image to be recognized by using a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels; a target label of the item is determined based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model. Namely, the embodiment of the invention improves the quality of the labeling image by preprocessing the labeling image, further improves the accuracy of label identification, performs label identification on the image to be identified through a label prediction network to obtain a plurality of predicted labels, and screens or determines the plurality of predicted labels by combining a label determination model to obtain a target label, thereby improving the efficiency and accuracy of identification, and avoiding the influence of the environment on the identification effect under the condition of no personnel participation.
Fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. FIG. 4 illustrates a block diagram of an exemplary electronic device 12 suitable for use in implementing embodiments of the present invention. The electronic device 12 shown in fig. 4 is only an example and should not bring any limitation to the function and the scope of use of the embodiment of the present invention.
As shown in FIG. 4, electronic device 12 is embodied in the form of a general purpose computing device. The components of electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.
The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30 and/or cache memory 32. The electronic device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 4, and commonly referred to as a "hard drive"). Although not shown in FIG. 4, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of the described embodiments of the invention.
The processing unit 16 executes various functional applications and data processing by executing programs stored in the system memory 28, for example, implementing a tag identification method provided by an embodiment of the present invention, the method including:
acquiring a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
determining, with a tag determination model, a target tag for the item based on the plurality of predicted tags and the confidence level of each predicted tag.
An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the tag identification method as described above, and the method includes:
acquiring a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
determining, with a tag determination model, a target tag for the item based on the plurality of predicted tags and the confidence level of each predicted tag.
Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, or the like, as well as conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A tag identification method, comprising:
acquiring a labeling image of an article, and preprocessing the labeling image to obtain an image to be identified;
predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
determining, with a tag determination model, a target tag for the item based on the plurality of predicted tags and the confidence level of each predicted tag.
2. The method according to claim 1, wherein the tag prediction network comprises a convolutional network and a long-short memory network, and the predicting the tags in the image to be recognized by using the tag prediction network to obtain the plurality of predicted tags in the image to be recognized and the confidence of each predicted tag in the plurality of predicted tags comprises:
extracting a characteristic vector of a detection area in the image to be identified by utilizing the convolutional network;
and extracting character features of the feature vector according to the semantic relation of the feature vector by using the long and short memory network, and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
3. The method according to claim 2, wherein the convolutional network comprises a convolutional layer and an anti-convolutional layer, and the extracting feature vectors of the detection regions in the image to be identified by using the convolutional network comprises:
inputting the image to be identified into a convolutional layer to extract image features in the image to be identified, and using the deconvolution layer to perform upsampling on the convolutional layer to obtain a convolutional image, wherein a correlation exists between the convolutional image and pixels of the image to be identified;
and carrying out pixel classification on the convolution image to obtain a two-dimensional thermodynamic diagram of the image to be identified, and carrying out image segmentation on the image to be identified according to the two-dimensional thermodynamic diagram to obtain a feature vector of the detection area.
4. The method according to claim 2, wherein the extracting, by using the long and short memory network, the text features of the feature vector according to the semantic relationship of the feature vector, and predicting the labels in the image to be recognized according to the text features to obtain a plurality of predicted labels in the image to be recognized and a confidence level of each predicted label in the plurality of predicted labels comprises:
inputting the feature vector of the detection area into the long and short memory network to obtain the semantic relation of the feature vector, and extracting the character features of the semantic relation;
and predicting the label in the image to be recognized according to the character features to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels.
5. The method of claim 1, wherein after determining the target label for the item based on the plurality of predicted labels and the confidence level of each predicted label using a label determination model, further comprising:
extracting attribute information from the target label of the article, and storing the attribute information and the labeling image in an information database in a correlation manner, so that the attribute information of the article is inquired according to the labeling image.
6. The method of claim 1, wherein the label determination model is obtained as follows:
obtaining a training sample, wherein the training sample comprises a plurality of identification images, and each identification image comprises a plurality of prediction labels and confidence degrees of each prediction label in the plurality of prediction labels;
screening a plurality of prediction labels in each recognition image in the training sample and the confidence coefficient of each prediction label in the plurality of prediction labels by using a time sequence class classification model to obtain a target label corresponding to each recognition image;
calculating a loss function according to the target label corresponding to each identification image and the actual label corresponding to each identification image;
and performing back propagation according to the loss function to optimize the time sequence class classification model to obtain the label determination model.
7. The method of claim 1, wherein after predicting the tag in the image to be recognized using a tag prediction network, further comprising:
and inputting the image to be identified into the label prediction network to obtain identification error information of the image to be identified, and performing fault early warning according to the identification error information.
8. A label identification device, comprising:
the image acquisition module is used for acquiring a labeling image of an article and preprocessing the labeling image to obtain an image to be identified;
the network prediction module is used for predicting the labels in the image to be recognized by utilizing a label prediction network to obtain a plurality of predicted labels in the image to be recognized and the confidence coefficient of each predicted label in the plurality of predicted labels;
a tag determination module to determine a target tag of the item based on the plurality of predicted tags and the confidence level of each predicted tag using a tag determination model.
9. An electronic device, characterized in that the electronic device comprises:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the tag identification method of any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the tag identification method according to any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210676549.7A CN114972880A (en) | 2022-06-15 | 2022-06-15 | Label identification method and device, electronic equipment and storage medium |
| PCT/CN2023/079023 WO2023241102A1 (en) | 2022-06-15 | 2023-03-01 | Label recognition method and apparatus, and electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210676549.7A CN114972880A (en) | 2022-06-15 | 2022-06-15 | Label identification method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114972880A true CN114972880A (en) | 2022-08-30 |
Family
ID=82963761
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210676549.7A Pending CN114972880A (en) | 2022-06-15 | 2022-06-15 | Label identification method and device, electronic equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114972880A (en) |
| WO (1) | WO2023241102A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542250A (en) * | 2023-06-29 | 2023-08-04 | 杭州同花顺数据开发有限公司 | Information extraction model acquisition method and system |
| WO2023241102A1 (en) * | 2022-06-15 | 2023-12-21 | 卡奥斯工业智能研究院(青岛)有限公司 | Label recognition method and apparatus, and electronic device and storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8774515B2 (en) * | 2011-04-20 | 2014-07-08 | Xerox Corporation | Learning structured prediction models for interactive image labeling |
| CN107886028A (en) * | 2016-09-29 | 2018-04-06 | 九阳股份有限公司 | The food materials input method and food materials input device of a kind of refrigerator |
| CN110827247B (en) * | 2019-10-28 | 2024-03-15 | 上海万物新生环保科技集团有限公司 | Label identification method and device |
| CN112036509A (en) * | 2020-09-30 | 2020-12-04 | 北京百度网讯科技有限公司 | Method and apparatus for training image recognition models |
| CN113420757B (en) * | 2021-08-23 | 2021-11-30 | 北京每日优鲜电子商务有限公司 | Text auditing method and device, electronic equipment and computer readable medium |
| CN114972880A (en) * | 2022-06-15 | 2022-08-30 | 卡奥斯工业智能研究院(青岛)有限公司 | Label identification method and device, electronic equipment and storage medium |
-
2022
- 2022-06-15 CN CN202210676549.7A patent/CN114972880A/en active Pending
-
2023
- 2023-03-01 WO PCT/CN2023/079023 patent/WO2023241102A1/en unknown
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023241102A1 (en) * | 2022-06-15 | 2023-12-21 | 卡奥斯工业智能研究院(青岛)有限公司 | Label recognition method and apparatus, and electronic device and storage medium |
| CN116542250A (en) * | 2023-06-29 | 2023-08-04 | 杭州同花顺数据开发有限公司 | Information extraction model acquisition method and system |
| CN116542250B (en) * | 2023-06-29 | 2024-04-19 | 杭州同花顺数据开发有限公司 | Information extraction model acquisition method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023241102A1 (en) | 2023-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109902622B (en) | Character detection and identification method for boarding check information verification | |
| CN114972880A (en) | Label identification method and device, electronic equipment and storage medium | |
| CN112613569B (en) | Image recognition method, training method and device for image classification model | |
| CN112070135A (en) | Power equipment image detection method and device, power equipment and storage medium | |
| CN112016638A (en) | Method, device and equipment for identifying steel bar cluster and storage medium | |
| CN108550054B (en) | Content quality evaluation method, device, equipment and medium | |
| CN109829458B (en) | Method for automatically generating log file for recording system operation behavior in real time | |
| CN110826494A (en) | Method and device for evaluating quality of labeled data, computer equipment and storage medium | |
| CN113963147B (en) | Key information extraction method and system based on semantic segmentation | |
| CN113012200B (en) | Method and device for positioning moving object, electronic equipment and storage medium | |
| CN112149663A (en) | RPA and AI combined image character extraction method and device and electronic equipment | |
| CN113239807B (en) | Method and device for training bill identification model and bill identification | |
| CN112232354A (en) | Character recognition method, device, equipment and storage medium | |
| CN114694130A (en) | Method and device for detecting telegraph poles and pole numbers along railway based on deep learning | |
| CA3035387C (en) | Digitization of industrial inspection sheets by inferring visual relations | |
| CN111126112B (en) | Candidate region determination method and device | |
| CN117437647B (en) | Oracle character detection method based on deep learning and computer vision | |
| CN110728287A (en) | Image recognition method and device, electronic equipment and storage medium | |
| CN113936232A (en) | Screen fragmentation identification method, device, equipment and storage medium | |
| CN114120086A (en) | Pavement disease recognition method, image processing model training method, device and electronic equipment | |
| CN111738290B (en) | Image detection method, model construction and training method, device, equipment and medium | |
| CN117593244A (en) | Film product defect detection method based on improved attention mechanism | |
| CN113762455A (en) | Detection model training method, single character detection method, device, equipment and medium | |
| CN116030050A (en) | On-line detection and segmentation method for surface defects of fan based on unmanned aerial vehicle and deep learning | |
| CN115546824B (en) | Taboo picture identification method, apparatus and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |