74results about How to "Quick annotation" patented technology

The present invention provides a unique system and method for facilitating browsing, sorting, clustering, and filtering any number of media objects grouped together in a quick and easy manner. In particular, the invention involves annotating the media objects, which have been clustered, sorted, and / or filtered en masse rather than individually, with metadata. The media objects may be selected by a user and annotated with a plurality of metadata as desired by a user in order to optimize utilization of the media objects. The media objects may also be combined and / or mixed in any number of ways to create new media objects. Media objects may include an audio component, a visual component, a portion thereof, and / or any combination thereof.

The present invention provides a unique system and method for facilitating browsing, sorting, clustering, and filtering any number of media objects grouped together in a quick and easy manner. In particular, the invention involves annotating the media objects, which have been clustered, sorted, and / or filtered en masse rather than individually, with metadata. The media objects may be selected by a user and annotated with a plurality of metadata as desired by a user in order to optimize utilization of the media objects. The media objects may also be combined and / or mixed in any number of ways to create new media objects. Media objects may include an audio component, a visual component, a portion thereof, and / or any combination thereof.

The present invention provides a unique system and method for facilitating browsing, sorting, clustering, and filtering any number of media objects grouped together in a quick and easy manner. In particular, the invention involves annotating the media objects, which have been clustered, sorted, and / or filtered en masse rather than individually, with metadata. The media objects may be selected by a user and annotated with a plurality of metadata as desired by a user in order to optimize utilization of the media objects. The media objects may also be combined and / or mixed in any number of ways to create new media objects. Media objects may include an audio component, a visual component, a portion thereof, and / or any combination thereof.

The present invention provides a unique system and method for facilitating browsing, sorting, clustering, and filtering any number of media objects grouped together in a quick and easy manner. In particular, the invention involves annotating the media objects, which have been clustered, sorted, and / or filtered en masse rather than individually, with metadata. The media objects may be selected by a user and annotated with a plurality of metadata as desired by a user in order to optimize utilization of the media objects. The media objects may also be combined and / or mixed in any number of ways to create new media objects. Media objects may include an audio component, a visual component, a portion thereof, and / or any combination thereof.

The present invention provides a unique system and method for facilitating browsing, sorting, clustering, and filtering any number of media objects grouped together in a quick and easy manner. In particular, the invention involves annotating the media objects, which have been clustered, sorted, and / or filtered en masse rather than individually, with metadata. The media objects may be selected by a user and annotated with a plurality of metadata as desired by a user in order to optimize utilization of the media objects. The media objects may also be combined and / or mixed in any number of ways to create new media objects. Media objects may include an audio component, a visual component, a portion thereof, and / or any combination thereof.

The present invention relates to methods and systems for generating annotated data for training vehicular driver assist (DA) and autonomous driving (AD) active safety (AS) functionalities and the like. More specifically, the present invention relates to methods and systems for the fast estimation of three-dimensional (3-D) bounding boxes and drivable surfaces using LIDAR point clouds and the like. These methods and systems provide fast and accurate annotation cluster pre-proposals on a minimally-supervised or unsupervised basis, segment drivable surfaces / ground planes in a bird's-eye-view (BEV) construct, and provide fast and accurate annotation cluster pre-proposal labels based on the feature-based detection of similar objects in already-annotated frames. The methods and systems minimize the expertise, time, and expense associated with the manual annotation of LIDAR point clouds and the like in the generation of annotated data for training machine learning (ML) algorithms and the like.

The invention discloses a pathological image automatic labeling method based on reinforcement learning and a deep neural network, and the method comprises the following steps: carrying out the multi-scale superpixel division of a dyed standardized original pathological image, and marking the epithelial probability threshold of superpixels; constructing a super-pixel classification training set; training the marked superpixels based on a machine learning method to obtain a learning model, classifying the superpixels in the test image by the learning model, awarding and punishing a classification result by a pathologist, feeding back awarding and punishing results to the learning model, and adjusting the learning model again through an awarding and punishing mechanism; constructing a training set of an end-to-end labeling model by adopting a pre-established method; adopting end-to-end learning based on a deep neural network to automatically mark and segment an organization region; constructing a labeling model; and testing the constructed annotation model by using pre-configured real data. The method has the beneficial effect that the learning models of different tissue regions can be quickly, accurately and automatically marked.

The invention relates to the technical field of power distribution networks, in particular to a rapid distribution line drawing method. The method includes the steps of S1, establishing a component document library, a parameter document and a layer initiating document; S2, calling the component document library, reading a component link list generated through component documents in the component document library, calling the parameter document, reading parameter information, containing an intelligent comparison table, of the parameter document, calling the layer initiating document, and loading layer information of the layer initiating document onto a corresponding layer generated in CAD; S3, calling the component documents of an entity object according to the component link list, instantiating the component documents into component symbols, inserting the component symbols into the corresponding layer in the CAD, starting a preset intelligent interface in drawing, and inserting pole and tower appurtenance between the component symbols in reference to the intelligent comparison table. By setting the intelligent interface, the pole and tower appurtenance is automatically generated in the drawing process, and the power distribution network drawing efficiency is improved.

The invention discloses a method for marking key points in human face images, and the method comprises the steps: pre-storing a plurality of human face images and the basic coordinates of preset key points in each human face image; loading and storing the human face images in advance before the human face images are displayed, searching the basic coordinates of preset key points in each human face image in advance, and storing the searched coordinates of preset key points in each human face image; displaying the human face images which are load and stored in advance, and carrying out the initial marking and displaying of the preset key points in each human face image according to the coordinates of the preset key points; presetting a plurality of image marking operation instructions and keyboard key triggering signals corresponding to the plurality of image marking operation instructions, receiving the triggering signals, obtaining the corresponding marking operation instructions, and carrying out the marking and storing of the preset key points in the human face images, the posture of a human face, and human face expressions according to the basic coordinates of the preset key points. The method and system achieves the quick and effective marking of the key points in the human face images while guaranteeing high quality, and improve the work efficiency of a user.

The invention relates to big data, the invention provides a webpage labeling method. The method comprises the following steps: receiving a labeling request; determining a webpage to be labeled according to the labeling request, monitoring a trigger event on the to-be-labeled webpage; obtaining a target area, extracting information in the target area, determining the type of the information; obtaining a plurality of preset tags corresponding to the type from a configuration tag library, acquiring data corresponding to the plurality of preset tags; according to the plurality of preset labels andthe data, obtaining the data, construction of a target model, and inputting the information into the target model, generating annotation information, improving the generation efficiency of the annotation information, establishing an identifier according to the target area and the annotation information, and storing the identifier, thereby not only realizing quick annotation of the webpage, but also avoiding annotation information coverage generated when different users annotate the same area. The invention also relates to a blockchain technology, and the annotation information is stored in the blockchain.

The embodiment of the invention discloses a sample data labeling method and device for machine learning. The method comprises the steps of obtaining laser point cloud data to be labeled; displaying the laser point cloud data in a two-dimensional overlook interface according to a first mapping relation between a two-dimensional overlook coordinate system and a three-dimensional coordinate system; obtaining a first labeling operation input by aiming at the laser point cloud data displayed on the two-dimensional overlook interface; determining a bounding box corresponding to the first labeling operation according to the three-dimensional coordinates of the laser point cloud data; determining marking information corresponding to the laser point cloud data according to the bounding box, whereinthe three-dimensional coordinate system is a coordinate system where the laser point cloud data is located, and the two-dimensional overlook interface corresponds to the two-dimensional overlook coordinate system. By applying the scheme provided by the embodiment of the invention, the efficiency of marking the laser point cloud data can be improved.

The invention discloses a high-resolution SAR (Synthetic Aperture Radar) image marking method based on a supervised topic model. The high-resolution SAR image marking method comprises the following steps of: S1, image segmentation which further comprises the processes of segmenting an SAR image into a plurality of sub images with the same sizes and dividing each sub image into mutually-nonsuperposed rectangular regions, wherein each rectangular region is called a word; S2, image representation: generating files by using the sub images and generating a collected works by using all the files; S3, knowledge input which further comprises the processes of selecting a part of sub images with typical scenes to artificially mark based on key words, wherein a collection formed by the key words is called an implicit category of the files; S4, training reasoning: finishing the primary marking of the SAR image; and S5, semantic analysis: carrying out semantic analysis on the marking result through firstly validating semantic knowledge according to the classification result of all the sub images and the marking result of the words so as to obtain the marking result conforming to human semantics.

The invention provides a satellite image ground object marking device and method; the satellite image ground object marking device comprises the following parts: a first unit configured to obtain a training set and extract ground object marking features of the training set, thus forming a training model, wherein the training set comprises a plurality of first satellite images with finished groundobject marking; a second unit configured to obtain the training model and marking ground objects in the second satellite image according to the training model. The satellite image ground object marking method comprises the following steps: extracting ground object marking features of the training set, forming the training model; marking the ground objects in the second satellite image according tothe training model. The satellite image ground object marking device and method can automatically mark the satellite image ground objects, can fast mark ground objects with high accuracy, and have low sharpness requirements for the satellite images.

The invention discloses a part machining method for automatically obtaining dimensional tolerance and belongs to the technical field of machining. According to the method, information in tolerance data is integrated in UG NX7.5 software, an accurate and fast tolerance information marking tool is provided for model information marking people by recognizing and judging model dimensions and retrieving the tolerance data, errors caused by manual operation are reduced, the part machining and production efficiency and quality are improved, and the project quality requirements are met. Besides, the method not only can complete accurate dimensional tolerance value entry, but also can adjust tolerance marking styles automatically according to dimensional tolerance types, and the dimensional tolerance marking styles are prevented from being repeatedly adjusted.

The invention relates to an agricultural non-point source pollution multi-source heterogeneous big data association method based on attribute classification. Compared with the prior art, the defect that efficient association is difficult to perform according to the data attributes is overcome. The method comprises the following steps of judging whether multi-source heterogeneous big data belongs to the quantitative data or the qualitative data; classifying the quantitative data by adopting a support vector machine, metric learning and other methods; obtaining the quantitative characteristics of the qualitative data by adopting a text semantic mining method, and classifying the qualitative data by adopting a support vector machine, a metric learning method and the like; encoding the classified result to realize the association of multi-source heterogeneous big data. The invention further provides an agricultural non-point source pollution big data supervision platform. According to thepresent invention, the attributes of the agricultural non-point source pollution multi-source heterogeneous big data are used as the classification bases, different processing methods are adopted forquantitative and qualitative data, the classification of the agricultural non-point source pollution multi-source heterogeneous big data is achieved, and the correlation is conducted by means of the generated tree structure soil pollution attribute codes.

The invention provides a picture marking method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: identifying a to-be-processed picture, and determining detection frame information corresponding to each single product in the to-be-processed picture; sorting the detection frames corresponding to the detection frame information, and determining single item identifiers corresponding to the sorted detection frames based on preset single item identifier arrangement information; and according to the information of each detection frame,cutting out each single-item picture carrying a single-item identifier from the to-be-processed picture. According to the method and the device, each single product in the to-be-processed picture isframed through the identified information of each detection frame; According to the technical scheme, the pictures to be processed are cut on the basis of the detection frames, a large number of single-item pictures can be obtained rapidly, the collection efficiency is greatly improved, the single-item identifiers of the single-item pictures can be directly obtained on the basis of the single-itemidentifiers corresponding to the detection frames, a large number of labeling tasks can be completed rapidly, the labor cost is saved, and the labeling efficiency is improved.

The invention relates to an automatic image labeling system and method for new retail. The system comprises a test unit configured to a to-be-annotated picture and used for automatically annotating the picture corresponding to the to-be-annotated picture, a cascade network test unit used for acquiring position information of a target and the label of the target in the to-be-annotated picture, a test unit used for checking whether a labeling result is Groud Truth or not, and a test unit used for obtaining the labeling result. An image segmentation model for a to-be-labeled target is trained, the to-be-annotated picture is shot, and a paper sheet written with numbers on a target of the to-be-labeled picture for automatic labeling is placed, position information of a target in the picture isobtained through an image segmentation network, whether errors are generated or not, digital information on the target is obtained through a deep learning network, the label of the target is confirmedthrough a corresponding relationship between numbers and labels after checking errors do not exist, and a. Csv format file is generated according to a filename of the to-be-annotated picture, position information of each target in the picture and the labels to finish automatic labeling.

The invention relates to a quickly marking method of a schematic circuit diagram. The method comprises the following steps: at first, a parameter module is manufactured by parameters needing to be marked, wherein the parameter module comprises parameter information and parameter port information, then related devices corresponding to the parameter module as well as the connection relations of the related devices in the schematic circuit diagram are deleted, next, the parameter module is added in the schematic circuit diagram, and finally, the connection information of the parameter module is added in the schematic circuit diagram to realize the quick marking to the schematic circuit diagram. Through the adoption of the quickly marking method, the hand-manipulated workload of a designer can be reduced, the marking efficiency of the schematic circuit diagram can be improved greatly, and the quickly marking method has good compatibility, and is simple and easy to operate.

The invention relates to the technical field of electrocardiosignal labeling, and discloses a recommendation and distribution method for electrocardiogram data annotations. The method comprises the following steps: establishing an electrocardiogram sample data set, and annotating each piece of electrocardiogram sample data in the electrocardiogram sample data set to obtain standard label data of all the electrocardiogram sample data; judging whether the annotating result of each expert on each piece of electrocardiogram sample data is accurate or not according to the standard label data, and counting the annotation accuracy of each expert on each disease type to obtain an accuracy matrix; with the electrocardiogram sample data as input data and the standard label data as output data, training a neural network to obtain a prediction model; predicting to-be-annotated electrocardiogram data according to the prediction model to obtain a predicted disease type of the electrocardiogram data;and performing expert recommendation according to the predicted disease type and the accuracy matrix, and distributing the to-be-annotated electrocardiogram data to recommended experts for annotating. The method and the system have the technical effect of accurate electrocardiogram data annotation and delivery.

The invention firstly discloses a deep learning image labeling system. The system comprises an image preprocessing module, an image annotation module and a network training module which are connectedin sequence; the invention further discloses the deep learning image labeling method. The method comprises the following steps that a user uploads video data and performs video frame taking; fuzzy detection and key frame extraction are carried out on the video frame-taking picture to obtain a to-be-labelled picture; part of the to-be-labelled pictures is manually labelled to generate a manual labeling result, and the manual labeling result is used for algorithm training; the trained algorithm is used to perform automatic pre-labeling on the remaining to-be-labelled pictures to generate a pre-labelled result; and a pre-annotation result is rechecked and corrected to generate a final labeling result. The system realizes automatic labeling of image and video data, avoids time and labor wasteof manual labeling, effectively improves the data labeling efficiency, and realizes quick labeling of massive data; Implementation steps are provided for automatic labeling of image and video data, and data labeling workload is greatly reduced.

The embodiment of the invention discloses a text labeling method and device and a storage medium. According to the embodiment of the invention, the reference image can be obtained; marking a pluralityof angular point coordinates of a target area in the reference image and text position coordinates in the target area; obtaining a target image, respectively marking a plurality of angular points ofa target area in the target image; and calculating a perspective transformation matrix according to the plurality of angular point coordinates in the reference image and the plurality of angular pointcoordinates in the target image, and calculating text position coordinates in the target image according to the perspective transformation matrix and the text position coordinates in the reference image. According to the embodiment of the invention, after the reference image is preprocessed, the perspective transformation matrix between the reference image and the target image is calculated, so that the text position in the target image is calculated, the text position in the image can be quickly recognized and labeled, and the text labeling efficiency is greatly improved.

The invention provides a rhythm phrase labeling method and device, and relates to the technical field of speech synthesis. According to the rhythm phrase labeling method and device, in the rhythm phrase labeling process, a rhythm phrase boundary of voice data is determined according toe PPGs and a fundamental frequency value of each audio frame in the voice data, so that manual labeling of the rhythm phrases of the voice data can be avoided, the rhythm phrase boundary of the voice data can be quickly determined, and the rhythm phrase labeling efficiency is improved. Moreover, the prosodic phrase boundary is determined according to the PPGs and the fundamental frequency value of the voice data, and is not interfered by people as the dominant opinion, so that the accuracy of the prosodic phrase boundary can be improved.

The invention relates to a food recognition method and device based on deep learning, and the method comprises the steps: obtaining a to-be-recognized image, carrying out the food recognition of the to-be-recognized image, and determining the to-be-recognized image containing food as a first recognition image; then, conducting food target detection on the first recognition image, determining position parameters of food in the first recognition image, wherein the position parameters comprise coordinates of the food and the length and width of the area where the food is located in the first recognition image; determining a second recognition image according to the first recognition image and the position parameter; and finally, carrying out food recognition on the second recognition image todetermine the name of the food. According to the scheme provided by the invention, the problem that a food recognition mode is relatively tedious can be solved.

The invention discloses a marking device for pipeline processing. The device comprises a supporting table, a first pipe clamping assembly, a second pipe clamping assembly and a marking assembly, wherein the first pipe clamping assembly, the second pipe clamping assembly and the marking assembly are arranged above the supporting table, and the first pipe clamping assembly and the second pipe clamping assembly are arranged on the two sides of the marking assembly respectively. The device is convenient to use, on one hand, pipelines can be stably clamped by the first pipe clamping assembly and the second pipe clamping assembly, meanwhile, the pipelines with different diameters can be clamped by retractable internal clamp blocks, and furthermore, the pipelines are convenient to move by means of rollers; on the other hand, the pipelines can be rapidly marked by the marking assembly, each assembly can be supported by the supporting table, marking operation is convenient, so that labor resources are greatly saved, and the marking accuracy is substantially improved.

The invention discloses an electronic map information labelling method and device and a terminal. The method comprises the following steps of: obtaining a labelling information set for representing semantic information of a to-be-labelled map element; determining a labelling position set corresponding to the labelling information set according to a contour of the to-be-labelled map element; determining a labelling form set of the labelling information set according to the labelling position set; and labelling the labelling information set according to the determined labelling form set. According to the method, the labelling form set is determined according to the labelling position set, so that labelling information of the to-be-labelled map element is taken as a whole to determine labelling forms, and the labelling information which is taken as a whole has a certain system consistency.

The invention discloses a continuous frame picture annotation method. The method comprises the following steps: generating an annotation box in response to an annotation operation of an annotator on a first frame picture in a continuous frame picture group for a target object; respectively generating the annotation boxes on the remaining pictures except the first frame picture in the continuous frame picture group; identifying the target object contained in at least part of the residual pictures; and according to the target object contained in the at least part of the pictures, adjusting the respective annotation boxes of the at least part of the pictures. In the whole process, an annotator only needs to manually complete annotation of the first picture, then the target object in the continuous frame picture group can be rapidly and accurately annotated based on automatic annotation box generation, automatic target object recognition and automatic adjustment, the annotation efficiency is improved, and the annotation time is shortened. And wrong labeling caused by excessive dependence on manpower is avoided.

The invention discloses a sequence labeling method based on deep learning and application. The method comprises the following steps: acquiring a to-be-labeled text, and preprocessing the text; performing text translation processing and error correction processing on the preprocessed text; constructing a rule dictionary and a regular expression, pre-annotating the text after error correction processing based on the rule dictionary and the regular expression, and outputting pre-annotated content; and inputting the pre-labeled text into a plurality of sequence labeling models based on deep learning, and calculating the optimal labeling content according to the weight values of the plurality of sequence labeling models for the output result. According to the method, the sequence labeling efficiency and accuracy can be improved.

The present invention relates to methods and systems for generating annotated data for training vehicular driver assist (DA) and autonomous driving (AD) active safety (AS) functionalities and the like. More specifically, the present invention relates to methods and systems for the fast estimation of three-dimensional (3-D) bounding boxes and drivable surfaces using LIDAR point clouds and the like. These methods and systems provide fast and accurate annotation cluster pre-proposals on a minimally-supervised or unsupervised basis, segment drivable surfaces / ground planes in a bird's-eye-view (BEV) construct, and provide fast and accurate annotation cluster pre-proposal labels based on the feature-based detection of similar objects in already-annotated frames. The methods and systems minimize the expertise, time, and expense associated with the manual annotation of LIDAR point clouds and the like in the generation of annotated data for training machine learning (ML) algorithms and the like.

The invention discloses an optimization method, a system, a computer storage medium and an electronic device for machine learning, wherein, the obtaining method comprises the following steps of: obtaining at least one learning data; identifying the learning data to obtain an identification result, the identification result comprising: a feature map of the learning data and / or a category to which the learning data belongs; when the recognition result satisfies a preset selection rule, the learning data corresponding to the recognition result being used as the optimization data of the machine learning model. The invention can pertinently obtain the data which can obviously improve the performance of the machine learning model.

The invention provides a multi-label learning based activity prediction method for antibacterial peptide. The method is that corresponding amino acid components corresponding peptide sequences are extracted, and then corresponding matrix features are acquired according to physical and chemical attribute encoding, so as to generate feature vectors of the peptide sequences; the feature vector of each peptide sequence comprises two parts, namely, the amino acid components, and the matrix features extracted on the basis of the physical and chemical attribute encoding; the least-squares multi-label learning algorithm is performed to calculating to obtain a minimum transformation matrix W, then each label output of a sample to be tested can be obtained through the transformation matrix W, and a predication label vector set can be acquired according to each label output, thus the activity of antibacterial peptide sequences can be fast and accurately predicated according to the similar label vector set, the shape feature of each angle of each peptide sequence can be acquired, and as a result, the antibacterial peptide activity can be fast, accurately and automatically marked.