Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

27results about How to "Improve recognition results" patented technology

Method for identifying two-dimensional drawing of transformer substation

The invention relates to an identification method for a two-dimensional drawing of a transformer substation, belongs to the technical field of image identification, and solves the problems of low identification accuracy and content missing of the existing two-dimensional drawing. Comprising the following steps: preprocessing an image of a two-dimensional drawing of a transformer substation, segmenting to obtain sub-images, and obtaining to-be-detected images and corresponding auxiliary images based on the sub-images; sequentially transmitting the to-be-detected images and the corresponding auxiliary images into an element detection model, and fusing to obtain an element frame and an element category; intercepting an element box image according to the element box and the element category, and identifying a first text box and first text content in the element box image; respectively covering the first textbox and the element box to obtain two text detection images, and transmitting the two text detection images into a textbox detection model to obtain a second textbox outside the element box; intercepting a second textbox image and transmitting the second textbox image into a text recognition model to recognize a second text content; and associating the first textbox and the second textbox with the element box to obtain an identification result. And efficient and accurate drawing recognition is realized.
Owner:STATE GRID SHANGHAI MUNICIPAL ELECTRIC POWER CO

Dimension speech emotion recognition method, computer device, medium and product

The invention discloses a dimension speech emotion recognition method, a computer device, a medium and a product. The method comprises the following steps: based on a to-be-recognized voice, obtaining a target voice pronunciation feature; extracting auditory modulation cochlea features from the to-be-recognized speech based on a preset extraction strategy; analyzing the to-be-recognized voice based on a preset model, and extracting pre-training model features; performing time sequence coding on the target voice pronunciation features, the auditory modulation cochlea features and the pre-training model features to form three types of coding features, and then performing weighting processing to obtain three types of weighting features; then performing multi-task identification to obtain discrete emotion identification information and dimension emotion initial identification information; and processing the two types of information based on a preset loss calculation strategy to obtain a dimension speech emotion target recognition result. According to the method, the gap between the representation space and the actual emotion space in the prior art can be effectively reduced, and an accurate and reliable emotion recognition result can be obtained through comprehensive and detailed analysis.
Owner:INST OF PSYCHOLOGY CHINESE ACADEMY OF SCI

Grassland human settlement space ecological safety pattern construction method

The invention relates to a grassland human settlement space ecological security pattern construction method comprising the following steps: 1) identifying an ecological source land, including ecological source land preliminary identification and ecological source land screening; 2) based on grassland living space production and life characteristics, ecological resistance surface identification is carried out; and 3) combining an ecological source land evaluation result and an ecological resistance surface identification result, simulating an ecological corridor, an ecological key point and an ecological obstacle point by adopting a circuit theoretical model, and constructing an ecological safety pattern as an identification result of an ecological space in the grassland living space. The grassland human settlement space ecological safety pattern construction method developed by the invention can adapt to the particularity of the grassland pastoral area human settlement space, and provides a thought for grassland pastoral area territorial space layout optimization and management and control partition delimitation.
Owner:INNER MONGOLIA UNIV OF TECH

A comprehensive identification method of tight sandstone fracture logging combined with deep learning

The present invention discloses a method for comprehensive identification of tight sandstone fractures by logging combined with deep learning, which is divided into the following four parts: preprocessing of logging data, including denoising of logging curves based on multi-scale wavelet transform and standardization of logging curves based on fast Fourier transform; lithology identification based on LightGBM framework, lithology constraints for subsequent identification results; comprehensive prediction of fractures combined with deep learning, including fracture identification based on comprehensive probability index method, fracture identification based on one-dimensional multi-scale wavelet decomposition, fracture identification based on GaussianNaive Bayes discriminant method and fracture identification based on CNN-LSTM model, to achieve comprehensive identification of fractures by multiple means; result output, weighting the comprehensive identification results and multiplying them with the lithology identification results to achieve lithology constraints, and outputting the comprehensive identification probability of tight sandstone fractures based on lithology constraints. The present invention adopts multiple methods to evaluate the degree of fracture development in parallel, and selects or weights the identification results to obtain better identification results.
Owner:CHINA UNIV OF GEOSCIENCES (WUHAN)

Classified playing method based on audio recognition, sound equipment, equipment and storable medium

The invention belongs to the technical field of audio detection, and provides a classified playing method based on audio recognition, sound equipment, equipment and a storable medium. The method comprises the steps of obtaining audio information; classifying the audio information based on a detection model composed of a CNN network, a long short-term memory network, multi-modal compact bilinear pooling and a classifier; according to the classification result, a corresponding playing mode is matched for the audio information, the loudspeaker is switched to the playing mode when playing the audio information, the playing mode comprises one of music, movies, human voices, sports, games and festivals, and when the playing mode is music, the music genre is determined according to the classification result, and a corresponding equalizer playing effect is selected; according to the invention, problems encountered in audio playing in the prior art can be solved.
Owner:HUIZHOU ZHONGZHE SHANGLANBO TECH CO LTD

Vehicle driving area detection method, electronic equipment and storage medium

The invention relates to a vehicle driving area detection method, electronic equipment and a storage medium. The method comprises the following steps: acquiring image data and point cloud data in a driving area acquired by a target vehicle; lane line detection is carried out on the driving area based on the image data, and obstacle detection is carried out on the driving area based on the point cloud data; and when it is detected that a lane line exists in the driving area, obtaining a detection result of whether an obstacle exists in a target lane corresponding to the lane line. According to the embodiment of the invention, the image data and the point cloud data are combined, lane line detection is carried out by utilizing the image data, and obstacle detection is carried out by utilizing the point cloud data, so that whether an obstacle exists in the target lane or not can be quickly and accurately judged, the computing power can be saved to a great extent, and a better recognition result can be obtained.
Owner:BEIJING TIEMUNIU INTELLIGENT MACHINERY TECHNOLOGY CO LTD

Vehicle machine voice recognition method and system

The invention relates to the technical field of voice recognition, in particular to a vehicle voice recognition method and system, and the method comprises the steps: collecting voice data in a cab through a sound pickup, and recording the voice data as initial voice data; detecting and analyzing the initial voice data by using a VAD algorithm, and identifying voice frames in the initial voice data; analyzing each voice frame to obtain defect voice data; repairing the defective voice data, and recording the defective voice data as corrected voice data; evaluating a restoration result based on the corrected voice data; by analyzing the defective voice data and generating the corrected voice data, the system can optimize the recognition result and reduce the risk of misoperation, and meanwhile, by judging the accuracy of the corrected voice data, whether to directly execute an instruction or return confirmation can be intelligently determined, the user experience and the operation safety are improved, and the user experience is improved. According to the method, improper operation caused by voice recognition errors can be reduced, and particularly in the driving process, safety can be better guaranteed.
Owner:SIMBA NETWORK TECH (NANJING) CO LTD

Front vehicle detection method based on deep learning ROI (Region of Interest) correlation thundersight fusion algorithm

The invention aims to provide a novel thunder-vision fusion method, so as to solve the problem that thunder-vision fusion is relatively tedious in a traditional mathematical method, and further improve the safety of an automatic driving system. The invention provides a front vehicle detection method based on deep learning, and the method employs the data fusion of a millimeter wave radar and a camera, and achieves the vehicle detection (Deep-ROI, DR) through ROI correlation. Pictures shot under different conditions are cut into images with the same size to create a dep-fusion data set, a new method for manufacturing a radar data set is provided, multi-dimensional information of an effective target is projected to pixel points, and matrix format data with the same size as the images are generated. A backbone network of the deep learning network comprises two pipelines which are used for processing radar data and image data respectively, and two parameter association credibility (R parameter) and environment characteristics (E parameter) are added on the basis of a Resnet model for fusion. And the constructed deep learning network module is trained by using multiple frames of images containing different light and weather conditions and radar data.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA

Method, device and equipment for recognizing image text in test sheet and storage medium

ActiveCN116453149BImprove recognition resultsImprove recognition efficiency
The present application relates to the field of digital medical treatment, and discloses a kind of identification method of image text in test sheet, comprising: obtaining the position information of different regions in test sheet image, convert the text information in different position regions in check sheet image into text feature vector, using the first network model trained by target detection algorithm, test sheet image is encoded into visual feature vector, each visual feature vector represents the visual feature in the preset image region in test sheet image, after text feature vector and visual feature vector are combined into multimodal sequence feature, linear network mapping is carried out to obtain the feature representing entity in test sheet image, using the feature representing entity in test sheet image, the text in test sheet image and the corresponding relationship of text are identified, and the text array contained in test sheet is obtained.The present application enriches text feature by the mode of multimodal feature, avoids introducing cascade error prediction, and improves the identification result of text in test sheet image.
Owner:PING AN TECH (SHENZHEN) CO LTD

CoTNet-based micro-expression recognition method driven by space-time features and application of CoTNet-based micro-expression recognition method

The invention relates to a CoTNet-based micro-expression recognition method under space-time feature driving and application thereof. The method comprises the following steps: extracting an optical flow diagram of a micro-expression sample by using an optical flow method; extracting global features and local features by adopting CoTNet based on a context attention mechanism; based on CASMEII, SAMM and SMIC data sets, a micro-expression recognition model is trained and verified for the preprocessed data through a leaving-one cross validation method, and the micro-expression recognition model is used for recognizing medium-high-risk passengers in civil aviation security. According to the method, the CoTNet network and the optical flow method are combined, the micro-expressions are recognized by analyzing facial movement, key information in a micro-expression sequence is captured more comprehensively, the accuracy and robustness of micro-expression recognition are improved, meanwhile, micro-expression recognition is achieved based on the CoTNet micro-expression recognition method under space-time feature driving, and the method has the advantages of being high in recognition efficiency, high in accuracy and high in robustness. And the real emotional state of the passenger is analyzed, the risk of carrying dangerous goods and making dangerous behaviors of the passenger is judged, and related departments are assisted in identifying medium-and-high-risk passengers.
Owner:CIVIL AVIATION UNIV OF CHINA

Microscopic image enhancement and restoration method and system for tumor pathology

The invention discloses a microscopic image enhancement and restoration method and system for tumor pathology, and relates to the technical field of medical imaging. Comprising the following steps: S1, acquiring a normalized value of a microscopic image, comparing the normalized value with a preset normalized value, and determining a final microscopic image; s2, determining an artifact area and an artifact type corresponding to the final microscopic image, and processing the artifact area; and S3, optimizing the microscopic image after the artifact region processing, and reconstructing a three-dimensional blood vessel structure in the microscopic image to obtain an optimized microscopic image. According to the invention, artifact identification is carried out on the preprocessed microscopic image through the artifact detection model, so that tiny noise existing in the microscopic image can be reduced, and the identification efficiency and the identification result can be improved, so that the artifact area in the microscopic image can be better identified, and the identification accuracy of the microscopic image is improved. And the recovery efficiency of the artifact region is further improved.
Owner:THE SIXTH AFFILIATED HOSPITAL OF SUN YAT SEN UNIV

Voice transcription acceleration method based on artificial intelligence

The invention relates to the field of voice transliteration, in particular to a voice transliteration acceleration method based on artificial intelligence, which comprises the following steps: (S1) acquiring original voice data, preprocessing the acquired original voice data, then enhancing the preprocessed original voice data, performing feature extraction on the enhanced voice data to obtain voice features, and performing voice recognition based on the voice features to obtain voice recognition data; an identification result is obtained; and (S2) according to an identification result, generating a preliminary transcription text, optimizing the preliminary transcription text through a self-adaptive dynamic text optimization algorithm to obtain an optimized transcription text, and optimizing the transcription efficiency through an optimization acceleration algorithm in the transcription process. According to the voice transcription acceleration method based on artificial intelligence disclosed by the invention, background noise and other interferences are reduced, and the accuracy and speed of final text writing are improved.
Owner:NAT COMP NETWORK & INFORMATION SECURITY MANAGEMENT CENT

Abnormal limb behavior detection method for autism spectrum disorder based on posture tracking

The method for detecting abnormal limb behaviors of autism spectrum disorder based on posture tracking belongs to the technical field of video behavior analysis. The present invention first pre-processes the collected children's diagnosis video, then uses the re-trained target detection algorithm YOLOv3 to extract the detection frame of all people in the video, cuts out a sub-image, uses the HRNet human posture estimation method to detect human joint points on the sub-image, and uses the OSNet pedestrian re-identification network to extract the apparent features of the sub-image for multi-target tracking, obtains the motion trajectory of the child's left and right shoulders, elbows, and wrists, then calculates the mean value and standard deviation of the motion trajectory, uses the random forest algorithm to perform leave-one-out cross-validation on the obtained features, and finally makes a comprehensive judgment on the classification result; the present invention detects the motion trajectory of the child's left and right shoulders, elbows, and wrists, which is more comprehensive than single joint detection information, extracts two types of features related to the motion speed for comprehensive judgment, thereby improving the detection accuracy.
Owner:JILIN UNIVERSITY

Cloud recognition method, device, electronic device and readable storage medium for remote sensing images

The present application discloses a cloud recognition method, device, electronic device and readable storage medium for remote sensing images. The method includes updating the Fmask cloud recognition algorithm applicable to multispectral sensors in advance according to the data preprocessing method and band information of the hyperspectral sensor to obtain a hyperspectral cloud recognition method; in response to the data input mode selection instruction, obtaining the target data to be identified from the hyperspectral remote sensing data. Based on the hyperspectral cloud recognition method, the target data to be identified is image preprocessed to obtain the apparent reflectance information; based on the apparent reflectance information, the hyperspectral cloud recognition method is called to perform cloud recognition and cloud shadow recognition respectively to obtain cloud recognition results and cloud shadow recognition results; based on the cloud recognition results and cloud shadow recognition results, the cloud information of the hyperspectral remote sensing data is determined, thereby effectively improving the cloud recognition accuracy of the hyperspectral images of domestic hyperspectral satellites.
Owner:AEROSPACE INFORMATION RES INST CAS

Spectral peak identification method, equipment, medium and product based on iterative algorithm

The present invention relates to a spectrum peak recognition method, equipment, medium and product based on an iterative algorithm. The method comprises the steps of: inputting original spectrum data, calculating the signal-to-noise ratio estimation value of each data point based on a mean iteration signal-to-noise ratio estimation algorithm; traversing all data points in the spectrum, searching for data points that meet the signal-to-noise ratio threshold setting and have a trend of first rising and then falling, and marking them as spectrum peak candidate points; judging whether the spectrum peak candidate points meet the parameters; if yes, retaining the spectrum peak candidate points and outputting them; otherwise, removing the spectrum peak candidate points; screening all spectrum peak candidate points to obtain spectrum peaks that meet the conditions in the original spectrum data. Using an iterative method to search, recalibrate and adjust the identified spectrum peaks can achieve the effect of refining the peak center when the peak boundaries overlap, and using the peak interval threshold parameter to screen useless peaks. The method can better identify the spectrum peaks of the mass spectrum.
Owner:GUOKE XINZHI (TIANJIN) TECH DEV CO LTD

A dual-channel person re-identification method based on attention twin network

ActiveCN117132937BImprove recognition resultsfully exploit the similaritiesInternal combustion piston enginesCharacter and pattern recognitionFeature vectorFeature extraction
The present invention provides a dual-channel pedestrian re-identification method based on an attention twin network, comprising: sending a library image and a sample image to be identified into a first twin network for significant feature extraction, and obtaining a shallow feature vector of each image; performing similarity measurement on the shallow feature vectors of the sample image and the library image, calculating the loss L1 of the first twin network, and obtaining k library images with the highest similarity according to the loss value; sending the k library images and the sample image into a second twin network to calculate their loss L2; ​​at the same time, inputting the sample image and the k library images into a bidirectional long short-term memory network in order of similarity to calculate the upper and lower semantic feature vectors and the mean square error loss L3, and the upper and lower semantic feature vectors are aggregated to obtain a picture sorting feature vector; fusing the loss L2 of the second twin network with the picture sorting feature vector and calculating the distance, and then performing similarity aggregation to output a list of pictures that are most similar to the sample image to be identified.
Owner:NORTHEASTERN UNIV FOSHAN GRADUATE SCHOOL OF INNOVATION +1

An optimized method for measuring the volume between intraocular implants and anterior segment structures.

ActiveCN115147477BImprove recognition resultsEliminate the interference of human factorsImage enhancementImage analysisImaging processingImage resolution
This invention relates to the field of computer image processing technology, and particularly to an optimized method for measuring the volume between intraocular implants and anterior segment structures. The invention utilizes anterior segment OCT grayscale images, preprocesses the OCT images using specific image processing methods, generates a high-resolution OCT grayscale image mask using a specific mask generation method, and combines mathematical methods to automatically identify, 3D reconstruct, and measure the volume between the posterior surface of the intraocular implant and the anterior surface of the lens in the image. This invention fills the gap in the fully automated measurement of the volume between intraocular implants and anterior segment structures based on anterior segment OCT grayscale image recognition, proposes a new image preprocessing workflow, optimizes the speed and accuracy of Mask RCNN for high-resolution anterior segment OCT grayscale image recognition, and simultaneously enables 3D reconstruction and automatic measurement of the spatial volume between the posterior surface of the intraocular implant and the anterior surface of the lens.
Owner:SVISION IMAGING LTD

Engine blade damage identification and reconstruction method based on Mask RCNN algorithm

PendingCN119068244AHigh precisionHigh-precision recognition resultsBiological models3D modellingAlgorithmMotion recovery
The invention discloses an engine blade damage identification and reconstruction method based on a Mask RCNN algorithm, and belongs to the technical field of engine blade damage detection, and the method comprises the steps: building an engine blade defect data set; constructing an engine blade damage identification model based on a Mask RCNN algorithm, and training and verifying the model; identifying the blade damage by using the verified engine blade damage identification model; and performing three-dimensional reconstruction on the blade damage. According to the method, based on the Mask RCNN algorithm, accurate identification of the blade damage type, size and position is realized, fine three-dimensional reconstruction of the blade and the damage thereof is completed by using the motion recovery structure, and the accuracy and efficiency of engine blade damage identification and reconstruction are improved.
Owner:SICHUAN UNIV

Traffic sign category determination method, apparatus, vehicle, and electronic device

The application discloses a traffic sign category determination method and device, a vehicle and an electronic device, and relates to the technical field of vehicles. The method comprises the following steps: acquiring a to-be-tested image, wherein the to-be-tested image is an image comprising a traffic sign; performing gray scale processing on the to-be-tested image to obtain a position image, wherein the position image is used for representing the position of the traffic sign in the to-be-tested image; and determining a recognition result according to the to-be-tested image and the position image, wherein the recognition result is used for representing the category of the traffic sign. The application solves the technical problem that, in the prior art, the category of a traffic sign is identified by using a traditional computer graphics method, a machine learning method and other deep learning algorithms, so that the neural network model is complex, the recognition rate is relatively low, and the classification effect is not ideal.
Owner:CHINA FAW CO LTD

Big data-based electricity larceny detection method and system

The invention provides an electricity stealing detection method and system based on big data, and the method comprises the steps: obtaining power utilization data and line loss data corresponding to a target user through a detection device, and extracting the power utilization credit data of the target user; wherein the power consumption data comprises actual power consumption data and theoretical power consumption data, and the line loss data is used for calculating a line loss growth rate; constructing a power utilization error curve according to the power utilization data, and constructing a line loss change curve according to the line loss data; and performing time registration on the electricity utilization error curve, the line loss change curve and the user credit data, extracting electricity larceny characteristics from a registration result, and inputting the electricity larceny characteristics into an electricity larceny detection model to perform electricity larceny identification. According to the invention, the technical problem that the accuracy of the difference between the primary side power consumption data and the secondary side power consumption data is low and the judgment precision of the electricity stealing behavior is restricted is solved.
Owner:国网安徽省电力有限公司营销服务中心 +1

Method and system for evaluating taste information based on specific digitalization of biological perception

The present invention discloses a method and system for evaluating taste information based on specific digitalization of biological perception, which uses a taste sensor chip based on different types of taste receptor cells, and includes the following steps: using basic taste substances to test the response of the taste sensor chip to different basic tastes (sour, sweet, bitter, and salty), finding a specific channel that responds to a single basic taste substance, using different concentrations of basic taste substances to calibrate the basic taste intensity detection range of the specific channel, and adding a test sample to identify the taste information. The present invention simulates the taste recognition process of mammals, and uses the characteristics of different types of taste receptor cells that specifically recognize a single taste substance to quickly and accurately evaluate the intensity of different basic tastes in the same test sample, thereby realizing the digital evaluation of different taste information in the sample. Compared with traditional taste recognition methods based on taste receptor cells, this method has high recognition specificity and accuracy.
Owner:ZHEJIANG UNIV +1

Audio and video dual-mode speech recognition method based on attention mechanism

The invention discloses an audio and video dual-mode speech recognition method based on an attention mechanism. The audio and video dual-mode speech recognition method comprises the following steps: S1, acquiring audio and video data; s2, extracting video features; s3, audio feature extraction; s4, after the audio features and the video features are coded, inputting the audio features and the video features into a multi-modal attention layer for fusion, and realizing alignment correction of the audio features and the video features through an attention mechanism; s5, decoding to obtain audio recognition text data; s6, decoding to obtain video identification text data; s7, acquiring a deep semantic relationship between the audio recognition text data and the video recognition text data based on a multi-head attention mechanism, and fusing the audio recognition text data and the video recognition text data based on semantic information; and S8, outputting to obtain a speech recognition result. Compared with the prior art, the problems that audio and video information processing is disordered, the recognition matching degree is low and the accuracy is not high in the voice recognition process are solved.
Owner:SUZHOU LANGJIETONG INTELLIGENT TECH

Vehicle loading and unloading point identification method and device, computer equipment and storage medium

The invention relates to the technical field of Internet of Vehicles, in particular to a vehicle loading and unloading point identification method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring vehicle driving data and parking data of the Internet of Vehicles; according to the vehicle driving data and the parking data of the Internet of Vehicles, constructing a single vehicle loading and unloading point database; obtaining a comprehensive loading and unloading point database according to the single vehicle loading and unloading point database of all vehicles in the Internet of Vehicles; acquiring newly added parking data; and according to the comprehensive loading and unloading point database, analyzing the newly added parking data to obtain a loading and unloading point identification result. By adopting the method, when the starting point and the ending point of loading and unloading are identified, the dependence on external equipment can be reduced, and the universality and the reliability are enhanced.
Owner:FAW JIEFANG AUTOMOTIVE CO

Large model updating method and device, equipment, storage medium and product

The invention provides a large model updating method and device, equipment, a storage medium and a product, and the method comprises the steps: obtaining an instruction set, the instruction set comprises an adjustment instruction and a label, the adjustment instruction is used for carrying out instruction adjustment on a preset large model, and the label is used for representing the adjustment expectation of the adjustment instruction on the preset large model; grading a plurality of tokens included in the label to obtain a plurality of grading results; performing weighted loss calculation on the plurality of grading results according to the target loss weight information to obtain a target loss value; and updating the preset large model according to the target loss value to obtain a target large model. According to the method, after the instruction set of the large model is adjusted, the plurality of tokens in the instruction set are graded, and the plurality of obtained grading results are subjected to weighted loss calculation to obtain the target loss value, so that the preset large model is updated according to the target loss value, and the large model is enabled to pay more attention to the importance degree and the association relationship of the tokens; therefore, the identification result of the large model is improved.
Owner:北京中关村科金技术有限公司

Speech recognition method, apparatus and device, and storage medium

PendingEP4401074A4Improve recognition resultsThe recognition result is accurateSpeech recognition
Provided in the present application are a speech recognition method, apparatus and device, and a storage medium. The method comprises: acquiring a speech feature of target mixed speech and a speaker feature of a specified speaker; taking the direction of tending to a target speech feature as an extraction direction, and according to the speech feature of the target mixed speech and a speaker feature of a target speaker, extracting a speech feature of the target speaker from the speech feature of the target mixed speech, so as to obtain an extracted speech feature of the target speaker, wherein the target speech feature is a speech feature that is used to obtain a speech recognition result consistent with real speech content of the target speaker; and acquiring a speech recognition result of the specified speaker according to an extracted speech feature of the specified speaker. By means of the speech recognition method provided in the present application, speech content of a specified speaker can be recognized more accurately from mixed speech including speech of the specified speaker, thereby achieving a better user experience.
Owner:UNIV OF SCI & TECH OF CHINA +1

Method and device for training integrity identification model and electronic equipment

The embodiment of the invention discloses a training method of an integrity identification model, a traffic acquisition integrity identification method and device and electronic equipment, and the training method of the integrity identification model comprises the steps: randomly screening out a part of data from a plurality of pieces of acquired traffic data; according to whether the screened data is complete to generate a label, expanding the screened data and the label thereof by adopting a mask mode; converting the expanded data and labels into corresponding vector representation through a natural language processing model; performing feature extraction on the vector representation based on a self-attention mechanism to obtain low-dimensional features; and training a neural network model to identify whether the data represented by the low-dimensional features are complete or not by taking the low-dimensional features as input and a dichotomy result as output so as to obtain a completeness identification model, and the dichotomy result is used for identifying that the data are complete or incomplete.
Owner:CHINA TOWER CO LTD

A data augmentation method, application and device for small sample oracle bone inscription recognition

The present invention provides a data augmentation method, application and device for small sample oracle bone inscription recognition, characterized in that it includes the following steps: serializing large sample data and small sample data to obtain serialized large sample data and serialized small sample data; erasing the serialized large sample data to obtain erased large sample data and complete large sample data, encoding the erased large sample data to obtain encoded erased large sample data; inputting the encoded erased large sample data into a feature extraction network to obtain deep features; using a prediction network to predict the erased large sample data to obtain predicted large sample data; training and updating a deep learning network to obtain an augmenter; randomly erasing the serialized small sample data to obtain erased small sample data, and encoding the erased small sample data to obtain encoded erased small sample data; inputting the encoded erased small sample data into an augmenter to predict to obtain predicted small sample data, and using the predicted small sample data as an augmented sample.
Owner:FUDAN UNIVERSITY