101results about How to "Improve labeling accuracy" patented technology

The invention relates to an indoor scene semantic annotation method based on RGB-D data. According to the method, a coarse-to-fine global recursion feedback semantic annotation framework based on the RGB-D data is built, in addition, the whole semantic annotation framework is divided into two major parts including the coarse-granularity region stage semantic label deduction and fine-granularity pixel stage semantic label refinement. The framework is different from the traditional region stage or pixel stage semantic annotation framework, the framework rebuilds the relationship between the coarse-granularity region stage semantic label deduction and the fine-granularity pixel stage semantic annotation, and a reasonable global recursion feedback mechanism is introduced, so that the coarse-granularity region stage semantic annotation result and the fine-granularity pixel level semantic annotation result realize the alternate iterative updating optimization. Through adopting the mode, the multi-mode information of different region layers in the scene images is better merged, and the general problem that an annotation base element is difficult to be properly selected in the traditional indoor scene semantic annotation scheme is solved to a certain degree.

The invention discloses a performance fragment marking method, a video playing method, a video playing device, and a video playing system, and belongs to the multimedia technology field. The performance fragment marking method comprises steps that a multimedia file corresponding to a role is acquired; according to the multimedia file, the characteristic of the role can be determined; a target video is decoded to acquire a data frame and a plating time stamp corresponding to the data frame; a target data frame matched with the characteristic of the role is identified in the data frame of the target video; according to the playing time stamp of the target data frame, performance fragment information corresponding to the role is marked automatically. According to the invention, a server can be used for the automatic marking of a plurality of target videos at the same time, and therefore problems of operation editors of low efficiency and poor accuracy caused by the fact that the operationeditors can only mark a small part of vides in a limited time period can be solved, and effect of marking the target video effectively in the limited time period can be achieved.

The invention relates to a streetscape semantic annotation method based on a convolutional neural network and a semantic transfer conjunctive model. A device according to the streetscape semantic annotation method comprises a deep characteristic extracting part and a soft limited semantic transfer part. A more balanced training set is constructed, and furthermore a super-pixel classification deep model with prior information is trained. According to the streetscape semantic annotation method, the prior information of a scene can be sufficiently mined, and a characteristic expression with more remarkable difference is learned so that the annotation accuracy of a superpixel is greatly improved. Through a Markov random field model, an initial result is optimized and an unnecessary noise is eliminated so that an annotation result is further improved. Finally per-pixel annotation accuracy and average classification accuracy are respectively higher than 77% and 53%.

The invention discloses an annotation data processing method and an annotation data processing system. The annotation data processing method comprises the steps that step S110: similarity of multiple annotation results related to annotation tasks is calculated; step S120: the similarity is compared with a similarity threshold, the process goes to step S130 if the similarity is greater than or equal to the similarity threshold, and the process goes to step S140 if the similarity is less than the similarity threshold; step S130: a situation that multiple annotation results pass quality detection is determined; and step S140: a situation that multiple annotation results do not pass quality detection is determined. According to the annotation data processing method and the annotation data processing system, the quality of the annotation results is automatically detected by utilizing the similarity so that annotation staff are enabled to possibly obtain the quality of the annotation results timely and then possibly correct annotation errors timely, and thus annotation accuracy can be effectively enhanced.

The invention discloses a relation extraction method in combination with clause-level remote supervision and semi-supervised ensemble learning. The method is specifically implemented by the following steps of 1, aligning a relation triple in a knowledge base to a corpus library through remote supervision, and establishing a relation instance set; 2, removing noise data in the relation instance set by using syntactic analysis-based clause identification; 3, extracting morphological features of relation instances, converting the morphological features into distributed representation vectors, and establishing a feature data set; and 4, selecting all positive example data and a small part of negative example data in the feature data set to form a labeled data set, forming an unlabelled data set by the rest of negative example data after label removal, and training a relation classifier by using a semi-supervised ensemble learning algorithm. According to the method, the relation extraction is carried out in combination with the clause identification, the remote supervision and the semi-supervised ensemble learning; and the method has wide application prospects in the fields of automatic question-answering system establishment, massive information processing, knowledge base automatic establishment, search engines, specific text mining and the like.

One or more embodiments of the invention provide an image data generation method and device. The method comprises the steps of obtaining a simulation object model of a target object and a simulation environment model of a target scene; constructing a simulation scene of the target scene based on the simulation object model and the simulation environment model; generating a rendered image based onthe simulation scene, and determining annotation information of the rendered image, the annotation information being used for representing distribution information of a simulation object model contained in the simulation scene in the rendered image. A simulation scene of a target scene is automatically constructed; rendering the simulation scene by using a three-dimensional rendering technology toobtain a plurality of target annotation images; therefore, a large number of actually shot images do not need to be shot on site, the actually shot images do not need to be labeled manually, the synthesized image with high image reality sense and high labeling accuracy can be generated quickly, and a large number of available sample data with labeling information are provided for model training.

The invention provides a labeling task distribution method and device, a computer storage medium and a computer device. The method comprises the steps of obtaining to-be-labeled task data associated with a labeling task distribution instruction and a labeling personnel list in response to the labeling task distribution instruction; determining task attribute information of the to-be-labeled task data; obtaining annotation capability attribute information corresponding to each annotation person in the annotation person list; and allocating a labeling task to each labeling person according to the labeling capability attribute information corresponding to each labeling person and the task attribute information. According to the technical scheme, the marking capacity of the marking personnel is considered when the marking tasks are distributed, the marking tasks matched with the marking capacity of the marking personnel are distributed to the marking personnel, and the marking accuracy ofthe marking personnel can be remarkably improved.

The invention discloses a method for automatically annotating remote sensing images on the basis of deep learning. The method for automatically annotating the remote sensing images includes extracting visual feature vectors of the to-be-annotated remote sensing images; inputting the visual feature vectors into a DBM (deep Boltzmann machine) model to automatically annotate the to-be-annotated remote sensing images. The DBM model implemented in the method sequentially comprises a visible layer, a first hidden layer, a second hidden layer and a tag layer from bottom to top, and is acquired by means of training. The method for automatically annotating the remote sensing images has the advantages that the deep Boltzmann machine model implemented in the method comprises the two hidden layers (namely, the first hidden layer and the second hidden layer respectively), accordingly, the problem of 'semantic gaps' in image semantic annotation procedures can be effectively solved by the two hidden layers, and the integral annotation accuracy can be improved.

The invention discloses a multi-source multi-view-angle transductive learning-based short video automatic tagging method. The method comprises the steps of obtaining short video data; preprocessing the short video data to generate image key frames consistent in format, audio tracks, texts and semantic tags; extracting multi-view-angle eigenvectors of the image key frames, the audio tracks and the texts; establishing a short video tagging database, wherein the multi-view-angle eigenvectors and the semantic tags are stored in the short video tagging database; calculating the similarity between the multi-view-angle eigenvectors; establishing a multi-view-angle fusion space through the similarity between the multi-view-angle eigenvectors; and transductively solving the multi-view-angle fusion space, and automatically tagging the semantic tags to the to-be-tagged short video data. The invention furthermore discloses a multi-source multi-view-angle transductive learning-based short video automatic tagging system. According to the method and the system, multi-source information attached to the short video data is fully considered, so that the tagging accuracy is improved.

The invention discloses a medical image marking method and a system thereof, wherein the method and the system belong to the field of medical image processing. The method comprises the steps of performing analysis and image preprocessing on the medical image which is uploaded by a user, and obtaining a processed medical image; according to a preset operation distribution strategy, distributing themarking operation of the processed medical image to an assigned user supplying a marking process which corresponds with the processed medical image to the assigned user, wherein the marking process is used for marking the processed medical image by the assigned user; acquiring the marking result of the processed medical image by the assigned user, and performing storage. The medical image markingmethod and the system thereof can perform optimized marking on the medical image, and particularly the multi-mode multi-sequence MR image, and furthermore realize flexible and quick professional marking on the medical image.

The invention provides a corpus text processing method and device and electronic equipment. The method comprises the steps of inputting a corpus text set to be processed into a language model, and obtaining feature vectors of corpus texts; performing clustering processing on the corpus text set based on a clustering algorithm and the feature vectors of the corpus texts to obtain corpus classification information; modifying intention category annotation information annotated by the target corpus text to obtain the target corpus text; and adding the target corpus text into the original trainingsample to train a language model to obtain an optimized language model. According to the invention, clustering processing is carried out on the corpus text set through the language model and the clustering algorithm, and the intention category annotation information annotated by the target corpus information in the corpus classification information is corrected to train the language model, so thatthe language model can be iteratively optimized in the use process, the generalization ability of the language model and the clustering algorithm is improved, and the labeling accuracy of the intention category labeling information corresponding to the corpus text is improved.

The invention discloses a collaborative filtering-based teaching video labeling method. The collaborative filtering-based teaching video labeling method mainly solves the shortcoming of the low accuracy of teaching video labeling in the prior art. The method is implemented through the steps of inputting a teaching video and performing caption key frame extraction on the teaching video according to captions to obtain D key frames; performing caption extraction on the D key frames through optical character software and performing text correction and deleting on obtain captions to obtain D text documents; performing shot segmentation on the teaching video by combining the D text documents with a Gibbs sampler to segment the teaching video into M shots; labeling a part of the M shots, computing the cosine similarity between the labeled shots and unlabeled shots through a collaborative filtering method, and selecting five words with the highest cosine similarity to label the unlabeled shots. The collaborative filtering-based teaching video labeling method takes the caption information in the teaching video into consideration, thereby effectively describing the teaching video, improving the labeling accuracy of the teaching video and being applicable to video teaching.

The invention relates to an interactive behavior recognition method and device, computer equipment and a storage medium. The method comprises the steps of obtaining a to-be-detected image; inputting the to-be-detected image into a preset multi-task model to obtain key points and a detection box of pedestrians in the to-be-detected image, the key points being located in the detection box, and the multi-task model being used for pedestrian detection and human body key point detection; According to the key points of the pedestrian and a preset article shelf image corresponding to the to-be-detected image, determining interaction behavior information of the pedestrian and the corresponding article shelf. By adopting the method, the interaction behavior of the pedestrian and the article can beefficiently identified.

The invention relates to artificial intelligence, and provides an OCR image sample generation method and device, a printing form verification method and device, equipment and a medium, and the methodcomprises the steps: receiving an image generation instruction, and obtaining an image sample; inputting the image sample into a preset font typesetting generation model, obtaining first annotation information by performing text detection and character recognition on the image sample, and obtaining a simulation result generated by reconstruction of the font typesetting generation model; inputtingthe image sample and the simulation image into a preset style synthesis model, extracting style features and content features by the style synthesis model, and generating a synthesis result by the style synthesis model; and obtaining an OCR image sample label, recording the synthesized image as an OCR image sample corresponding to the image sample, and associating the OCR image sample with the OCRimage sample label. In addition, the invention also relates to a blockchain technology, and the information can be stored in the blockchain node. According to the invention, the OCR image sample withthe same texture style as the image sample is automatically generated, and the sample label is automatically labeled.

The invention belongs to a digital image multi-semantic annotation method which is characterized by comprising the following steps in sequence: (1) inputting a plurality of digital images with known semantemes and all digital images to be annotated into a computer; (2) acquiring a characteristic vector set of all images by extracting characteristics; (3) establishing a mark vector of marked images and a final mark vector set of all images; (4) calculating a Gram matrix of the characteristic vector set; (5) acquiring a measurement value of the dependency degree of the characteristic vector set and the mark vector set by using a spatial dependency measurement method; (6) gradually increasing the dependency measurement value to the maximum in the iterative process, thereby obtaining confidence values that the images to be annotated belong to semantemes; and (7) setting a threshold, and judging the semantemes of the images to be annotated. The digital image multi-semantic annotation method has the advantages that firstly, the annotation effect can be improved by adopting a great number of images which are not semantically annotated, secondly, the method is applicable to the situation of multi-semantic annotation situation, and thirdly, the calculation speed is relatively high.

The invention relates to an image analysis method and a device, an electronic device and a storage medium, which relate to the technical field of image processing. The method comprises the following steps: scanning an object to be measured to obtain a sample image of the object to be measured; labeling a first type image in the sample image to obtain the label of the first type image, and generating an identification model according to the first type image and the label of the first type image; automatically labeling the second class of images other than a first class of images in the sample image by the recognition model to obtain the label of the second class of images. The present disclosure can realize automatic annotation and improve image annotation efficiency.

The invention discloses a data labeling method and device, and the method comprises the following steps: carrying out the alignment labeling of wake-up word data, and obtaining initial labeling data;training based on the initial labeling data to obtain a first wake-up model; traversing awakening word data which is not subjected to alignment labeling based on the first wake-up model, and determining an awakening word data set which can be successfully awakened; and performing alignment labeling on the wake-up words in the wake-up word data set based on the first wake-up model. According to thedata labeling method and device, the wake-up model is trained based on the initial labeling data, and then the matched wake-up words are labeled based on the special wake-up model, so that the labeling accuracy of the wake-up words is improved.

The invention discloses a multifunctional digital meter automatic reading method based on a target detection algorithm, and the method comprises the following steps: S1, carrying out the lightweight improvement of a YOLOv4 network model, and obtaining a YOLOv4 lightweight network model; S2, collecting the multifunctional digital table picture, and performing position calibration and category calibration on numbers on the multifunctional digital table picture; and S3, training the YOLOv4 lightweight network model by using the multifunctional digital table picture calibrated in the step S2 and the calibration information of the multifunctional digital table picture to obtain a YOLOv4 lightweight digital detection model. The multifunctional digital meter automatic reading method based on thetarget detection algorithm is high in recognition speed, high in recognition accuracy and good in stability, and the problems that an existing digital meter automatic reading method is prone to beingaffected by external environment changes and image imaging quality and lacks stability and accuracy of digital recognition are solved.

The invention discloses an iterative construction method and device for a military scenario text event extraction corpus. The method comprises the following steps of 1, preprocessing, and obtaining anoriginal data set represented by a word sequence; 2, constructing a seed data set, defining an event template, constructing an event trigger word dictionary, forming the seed data set through manualannotation, and dividing the seed data set into a seed training set and a test set; 3, training a model, training a machine learning model by using the seed training set, testing the model by using the test set, and optimizing the model parameters according to a test result to obtain a first learning model; 4, selecting an unlabeled training corpus, and inputting the unlabeled training corpus intothe first learning model to obtain a prediction result set; 5, correcting the prediction result set to form a new annotation corpus; and 6, through the continuous iteration, generating the training sets in sequence to form the event extraction corpus. According to the iterative construction method for the military scenario text event extraction corpus, the corpus construction efficiency is improved, the manual annotation cost is reduced, and the relatively higher corpus annotation accuracy is obtained.

The invention relates to an image labeling method and device, computer equipment and a readable storage medium. The method can comprise the following steps: marking the same to-be-marked image according to a pre-trained odd number of image marking models, and obtaining corresponding odd number of first marking images; Wherein the image labeling model is used for labeling a target object in an image; And according to the odd number of first annotation images, determining a target annotation image of the to-be-annotated image. Image labeling method, obtaining An odd number of first annotation images according to an odd number of pre-trained image annotation models; improving the marking efficiency of the to-be-marked image; Besides, the computer equipment determines the target annotation image of the to-be-annotated image according to the obtained odd number of first annotation images, so that the problem that the annotation quality of the to-be-annotated image is low due to insufficientpersonal experience of annotators or incomplete existing rules is avoided, and the annotation accuracy of the determined target annotation image is improved.

The invention discloses a text matching method and device, a computer system and a readable storage medium, which are suitable for the field of artificial intelligence, and comprise the following steps: identifying and obtaining a standard file, obtaining standard short sentences matched with standard regular expressions pre-stored in a database in the standard file, and summarizing to form a matched short sentence set; obtaining a clause element set according to the matching short sentence set and a clause regular expression pre-stored in a database library; splitting the current clause template and obtaining a short sentence set of the current clause template; recognizing clause short sentences in the current clause template short sentence set by utilizing the clause element set, and obtaining a specified short sentence set; and marking the current clause template according to the specified short sentence set to obtain a latest clause template. According to the method, the operationburden of the system is reduced, the annotation efficiency is improved, the association between the contract terms and the standard files is reflected to the maximum extent, and the annotation accuracy is improved.

The invention provides a text data labeling method and device, electronic equipment and a storage medium, relates to the technical field of data processing. The text data annotation method comprises the steps of obtaining to-be-annotated text data, and performing conversion processing on the text data according to a pre-trained topic model to determine vector representation data corresponding to the text data; determining the similarity between the text data through the vector representation data; determining similar text data of which the similarity exceeds a preset threshold, and extractingfirst text data and second text data of the similar text data in a preset similarity interval; and presenting the first text data and the second text data to a display interface, so that a target object marks the similar text data according to the first text data and the second text data. According to the method, the label labeling efficiency of the sample text data can be improved, and the user experience is improved.

The invention relates to the field of intelligent decision, and discloses an image classification method, which comprises the following steps: training a pre-constructed first convolutional neural network model by using a first annotated image set to obtain a first image classification model; performing image screening and segmentation processing on the to-be-labeled image set to obtain a segmented image set; performing classification labeling on the segmented image set by using a first image classification model to obtain a second labeled image set; combining the first annotated image set and the second annotated image set to obtain an annotated image set; using the annotation image set to carry out iteration annotation training on a pre-constructed second convolutional neural network model to obtain a target image classification model; and classifying the to-be-classified image by using the target image classification model to obtain a classification result. The invention also relates to a block chain technology, and the annotated image set can be stored in a block chain node. The invention further provides an image classification device, electronic equipment and a storage medium. According to the invention, the accuracy of image classification can be improved.

The invention discloses a labeling model training method and device, and relates to the field of computers, in particular to the field of data processing. The specific implementation scheme is as follows: training an initial labeling model by utilizing a first sample with a reference label to obtain a first labeling model; labeling the second sample by utilizing the first labeling model to obtaina second sample with model labeling; obtaining a third sample with a reference label, wherein the third sample is a part of the second sample; and optimizing the first annotation model according to the second sample with the model annotation and the third sample with the reference annotation. The data volume required by model training can be reduced, so that the manual operation amount and the time cost are saved.

The invention discloses a speech recognition method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of extracting pre-training features corresponding to an unlabeled first audio data sample through a feature extraction network, and obtaining a normalized weight vector of phonemes of the first audio data sample through a feature mapping network based on the pre-training features; and taking the normalized weight vector as a training target corresponding to the first audio data sample and taking a label of a labeled second audio data sample as a training target corresponding to the second audio data sample, training a speech recognition model, and performing speech recognition by using the trained speech recognition model. According to the implementation mode, the problems of data dependence and voice representation of voice recognition can be solved, the voice recognition performance is improved by effectively utilizing unmarked audio data in a voice recognition product, the manual marking cost is reduced, and the problems that in the prior art, voice phase information is ignored, and complex voice characteristic modeling capacity has defects are solved.

The invention relates to an obstacle map construction method, a cleaning robot and a storage medium. The obstacle map construction method comprises the steps of acquiring image information in the advancing direction of the cleaning robot; under the condition that it is judged that a target obstacle exists in the advancing direction of the cleaning robot according to the image information, determining position information of the target obstacle; and adding the position information into a grating map of the clean area to construct an obstacle map. According to the invention, the problems of tedious operation, poor experience and low labeling accuracy caused by manual labeling and adding of obstacle semantics during construction of an obstacle map in the prior art can be solved.

The embodiment of the invention discloses a medical image annotation method and device, equipment and a medium. The method comprises the steps of obtaining a to-be-annotated auxiliary image, inputtingthe to-be-annotated auxiliary image to a pre-trained target auxiliary image annotation model, obtaining an auxiliary annotation result output by the target auxiliary image annotation model; obtaininga to-be-annotated target image corresponding to the to-be-annotated auxiliary image, wherein the imaging mode of the to-be-annotated auxiliary image is different from the imaging mode of the to-be-annotated target image; and annotating the to-be-annotated target image based on the auxiliary annotation result to obtain a target annotation result of the to-be-annotated target image. According to the medical image annotation method provided by the embodiment of the invention, the to-be-annotated target image is annotated according to the auxiliary annotation structure of the to-be-annotated auxiliary image, so that the annotation accuracy of the target image is improved.

The invention relates to a digital labeling method and device for a three-dimensional object, belongs to the technical field of artificial intelligence, and solves the problems of high difficulty in knowledge labeling of the three-dimensional object and low labeling accuracy in the prior art. The method comprises the following steps: acquiring two-dimensional pictures, pose information and labeling labels of a three-dimensional object at different angles to form a training sample set; constructing an offset matrix and an enhanced angle rotation matrix of the sample based on the pose information in each sample; training the annotation network model by using the training sample set and the corresponding offset matrix and the enhanced angle rotation matrix to obtain an optimized annotation network model; and marking a to-be-marked three-dimensional object by using the optimized marking network model based on the two-dimensional picture and the pose information of the to-be-marked three-dimensional object. According to the method, the three-dimensional object labeling is converted into the two-dimensional labeling, the labeling difficulty is reduced, the association relationship between the labeling and the three-dimensional object is established through the artificial intelligence model, and the labeling accuracy of the three-dimensional object is improved.

The invention provides a text classification method and device and electronic equipment. The method comprises the following steps: inputting a to-be-classified text into a trained text classification model, and obtaining a text category of the to-be-classified text; the training mode of the text classification model is as follows: determining a plurality of text categories and an attribute rule of each text category based on the text data of which the statistical frequency is higher than a preset threshold value and / or the text data of which the semantic similarity meets a preset condition; and based on the determined text category and the attribute rule of the text category, labeling a plurality of sample texts, and training the initial model based on the plurality of sample texts carrying labeling information to obtain a text classification model. According to the mode, the text category and the attribute rule of the text category are obtained through manual summarization according to a small amount of selected representative unannotated text data, and then the text is automatically annotated according to the summarized rule, so that the annotated text with relatively high labeling accuracy is obtained; therefore, the classification accuracy of the text classification model obtained by training according to the annotated text is relatively high.

The invention discloses a construction drawing label generation method, a device, a system and a computer readable storage medium. The method comprises the following steps: determining each to-be-annotated primitive from a to-be-annotated construction drawing; obtaining the primitive information corresponding to each primitive to be labeled; according to a preset lead mark type and the information of each primitive, generating a lead mark corresponding to each to-be-marked primitive; displaying the marks with the leads in the construction drawing according to a preset rule so as to mark the corresponding primitives to be marked; according to the method, automatic marking of the primitives can be realized, the generation efficiency of the marking with the lead is improved, the marking accuracy is improved, the integrity, the accuracy and the usability of the construction drawing are improved, and accurate reference is provided for subsequent construction.