38results about How to "Increase labeling speed" patented technology

The invention discloses a machine learning method, device, equipment and system. The system comprises a processing module which is used for circularly executing a target operation until a test resultassociated with data in a test data set satisfies a preset condition. The processing module comprises a labeling module, a training module and a computing module. The storage module is used for storing the acquired data set and the test data set; The labeling module is used for labeling the data if the data in the data set is not labeled. The training module is used for carrying out feature extraction on the labeled data which is greater than or equal to a first preset threshold value, and inputting the extracted features into a pre-training model to obtain a trained training model; And the calculation module is used for inputting the data in the test data set into the trained training model, obtaining a test result associated with the data in the test data set, and assessing the test result to assess whether the test result meets a preset condition or not. By adopting the method, the model training speed can be improved.

The invention relates to a souvenir bag labeling machine. A souvenir bag conveying device is mounted on a rack and provided with a photoelectric sensor. A braking mechanism is mounted on the souvenirbag conveying device, and a drawing mechanism is mounted blow the front portion of the braking mechanism. A label peeling mechanism and a pneumatic label covering mechanism are mounted below the frontportion of the drawing mechanism, and a label flattening roller is mounted at the front end of the pneumatic label covering mechanism. A souvenir bag compressing device is mounted between the label peeling mechanism and the souvenir bag conveying device, and a wax paper recycling mechanism is mounted below the drawing mechanism. Wax paper separated from a label is recycled and wound around a label rolling device of the wax paper recycling mechanism through a drawing shaft and a pressing roll of the drawing mechanism. According to the souvenir bag labeling machine, the labeling work of souvenir bags can be automatically completed, and a synchronous contact pasting method of a brush-pasting method is adopted for achieving the aim of reliable and efficient labeling; and the whole mechanism is high in movement speed, simple in structure and convenient to operate.

The invention provides an image semantic segmentation method based on positive sample and label-free sample learning. The invention belongs to the technical field of computer vision, the method comprises a data preparation step, a data preprocessing step, a deep convolutional neural network construction step, a PU-Learning-based loss function design step, a loss function optimization learning step, and an iterative execution training step until a training result of an image semantic segmentation model satisfies a predetermined convergence condition. According to the invention, a deep neural network is adopted to extract to-be-segmented image features; on the basis, the invention further discloses a preparation method, according to the invention, a cross entropy loss function based on PU-Learning is designed; according to the method, the semantic segmentation model can be trained and optimized under the condition that only part of pixel levels are labeled, the semantic segmentation model can be trained and optimized in an end-to-end mode, meanwhile, direct supervision of the pixel levels is reserved to a certain extent, and the data labeling speed is increased while the good semantic segmentation quality is guaranteed.

The invention provides an interactive image text labeling method, which comprises the following steps: a server end receives an image and/or text data sample labeled by a user, and trains a model at the server end based on the sample. The trained model pre-annotates data which is newly loaded to the server end. The pre-annotated data is sent to the client for correction; And the server receives the corrected data to iteratively update the training model. The invention invokes the depth learning algorithm, can carry out on-line or off-line training on the labeled data samples to update the labeling model, continuously optimizes the labeling process, greatly improves the labeling efficiency, and saves human resources and time cost for customers.

The invention discloses a text labeling method of a streetscape video, terminal equipment and a storage medium. The method includes: acquiring a current frame in the streetscape video according to a clicking instruction, and adding a labeling point into the current frame according to the clicking position; generating the indicating boundary of a streetscape object according to the area where the labeling point is located; receiving text parameters associated with the labeling point; reading the frame, associated with the current frame, in the streetscape video, and matching the text parametersof the labeling point when the streetscape object exists in the frame associated with the current frame. The method has the advantages the streetscape video is output frame by frame, tensile deformation is avoided, the problem of unsmooth connection among images is avoided, and the display effect is improved; in addition, the labeling point and associated corresponding text parameters can be dynamically added to the streetscape object in the current frame, the labeling point and text parameters are matched with the corresponding streetscape object in the frame associated with the current frame, and labeling speed and streetscape object coverage degree are increased.

The invention discloses an image marking method for a streetscape video, a terminal device and a storage medium. The method comprises the steps of obtaining a current frame through a clicking instruction in the streetscape video, and adding a marking point to the current frame according to a clicking position; according to a region where the marking point is located, generating an indication boundary of a streetscape object; receiving image information associated to the marking point; and in the streetscape video, reading an associated frame of the current frame, and when the streetscape object exists in the associated frame, matching the marking point and the image information. The streetscape video is output frame by frame, so that the tensile deformation is not required, the problem ofunsmooth linkage between images does not exist, and the display effect is improved. Furthermore, the marking point and the image information associated with the marking point are dynamically added tothe streetscape object in the current frame, and the marking point and the image information are matched to the corresponding streetscape object in the associated frame, so that the marking speed andthe streetscape object coverage degree are improved.

The invention relates to data processing, and provides a collection call label establishment method and device, electronic equipment and a medium. The method is applied to the field of artificial intelligence. Acquiring target time and extracting a collection call audio; converting the collection call audio into a collection text and preprocessing the collection text; obtaining multiple segmentedwords, inputting the plurality of segmented words into a pre-trained first model, obtaining a first tag, selecting a pre-trained second model according to the first label, inputting the plurality of segmented words into a second model, obtaining a second tag, calculating an overdue rate and a confidence interval of the second label; and when it is detected that the overdue rate is within the confidence interval, the second label is determined as the target label of the call audio collection, so that the accuracy of the target label can be ensured without interference of human factors, the target label can be effectively applied, and the marking speed of the call audio collection can be increased. In addition, the invention also relates to a blockchain technology, and the target label can be stored in the blockchain.

The invention provides a point cloud data labeling method and device, and the method comprises the steps: firstly carrying out the object recognition of to-be-recognized point cloud data, and obtaining a detection box of an object in the to-be-recognized point cloud data; determining point cloud data to be labeled according to a detection box of an object identified in the to-be-recognized point cloud data; obtaining a manual labeling box of an object in the point cloud data to be labeled; and finally, according to the detection box and the manual labeling box, determining a labeling box of anobject in the to-be-recognized point cloud data. According to the invention, the detection box of an object obtained by automatically labeling point cloud data is combined with the residual point cloud data after the automatic point cloud data is labeled manually and the obtained manual labeling box for processing, so that the labeling box of the object can be accurately determined, the labelingspeed is increased and the labeling cost is reduced.

The invention discloses a marking system capable of image amplification. The marking system comprises a control module, and a storage module, a display mark module, a section module and an amplification module in communication connection with the control module; the storage module stores image data to be marked; the selection module calls the image data from the storage module and sends the data to the display mark module according to a selection instruction; the display mark module displays an image to be marked and receives a marking instruction from a marking staff; the control module controls the display mark module to display marking area selected by the marking staff in the image to be marked according to the marking instruction; the amplification module amplifies the marking area; and according to amplification by the amplification module, the control module controls the display mark module to further display the amplified marking area so that the marking staff can implement secondary marking adjustment on a marking object. Thus, a marking result is more accurate.

The invention discloses an image annotation method. The method comprises the following steps: determining at least one image block in a to-be-annotated image; respectively determining feature information of each image block, wherein the feature information is used for uniquely representing the corresponding image block; when target feature information matched with reference feature information exists in the feature information, performing annotation on the image block corresponding to the target feature information, so that labeling of a labeling object in the to-be-labeled image is achieved, and the reference feature information corresponds to the labeling object. According to the image annotation mode, the image blocks are automatically annotated, so that the problem that in the prior art, manual annotation is low in efficiency is solved, and a foundation is laid for efficiently obtaining a training sample.

The invention provides a method and a system for discovering a media account in a specific region on a social platform. An inventor discovers that high-quality seed media account information is difficult to quickly obtain by an existing method when a media account in a specific region on a social platform discovers a task, thus a set of method for quickly labeling a seed media account is designed, and the method comprises the steps of determining candidate accounts and classifying the accounts. The inventor finds that the feature design of the existing method has the problems of weak features, incapability of reflecting the features of the media account, difficulty in large-scale implementation and the like, so that the feature extraction method aiming at the features of the media account is designed. The method is high in speed and strong in extracted features. The inventor finds that the existing method is difficult to expand from the seed account to more media accounts in a high-efficiency and high-quality manner. Therefore, on the basis of solving seed account marking and feature extraction methods, two complementary expansion methods are designed to obtain high-quality and high-quantity media accounts.

The embodiment of the invention relates to the technical field of automatic driving, and provides an object recognition method and device and a storage medium, and the scheme comprises the steps: determining a target point cloud object in an initial point cloud according to a preset mapping relation and a target object in a first image, the preset mapping relationship comprises a coordinate mapping relationship between the first image and the initial point cloud, the target object and the target point cloud object correspond to the same entity target, and the first image and the initial point cloud are synchronized in a time domain; obtaining a target point cloud covered by the target object; determining a labeling sample of the target object according to the object type of the target object and the target point cloud; and outputting a recognition result, wherein the recognition result is obtained according to the labeling sample and the object type. According to the scheme, the recognition efficiency of the object in the point cloud data can be improved.

The invention belongs to the technical field of license plate detection, and relates to a weak supervision video license plate recognition method based on point labeling information of an eye tracker, which comprises the following steps of: generating a viewpoint position coordinate through the eye tracker, generating viewpoint region information from the coordinate position of the eye tracker by adopting a viewpoint smoothing mode, generating a large number of peripheral bounding box recommendations through a selective search algorithm, and carrying out point labeling on the peripheral bounding box recommendations. The consistency between the viewpoint area and the peripheral bounding box is calculated to determine an initial peripheral bounding box, and the quality of the initial peripheral bounding box is improved through inter-frame consistency; then, based on the viewpoint area information, a high-quality license plate peripheral bounding box is generated by adopting an area generation algorithm, the generated peripheral bounding box is used for training a test network, and a license plate detection result is generated by using the test network, so that the problems of inclination and distortion of the license plate are solved, and weak supervision video license plate detection based on the viewpoint area can be realized; and the method can also be used for various inclined target detection tasks such as scene text detection and supermarket commodity detection.

The invention discloses a text data annotation method. The method comprises the steps of text information extraction: determining a text search range, text annotation data and a text annotation standard; and text information segmentation and numbering: segmenting the text annotation data, and numbering the segmented data according to a segmentation sequence. According to the text data annotation method, text annotation data or data in a text search range are segmented before annotation, then the segmented text annotation data or data in the text search range are issued to different annotationplatforms for synchronous annotation, and finally annotated results are summarized, so that the text data annotation speed can be increased, the text data annotation time is shortened, and the pressure of the annotation platforms can be reduced. According to the invention, the text annotation data are divided, combined and then issued to different annotation platforms, and the data received by each annotation platform is incomplete, so that the risk of data leakage is reduced, and the security of text data annotation is improved.

The invention provides an image marking method and system, and the method comprises the steps: monitoring a frame selection instruction of a to-be-annotated image, so as to determine a region of interest; obtaining a minimum pixel value and a maximum pixel value in the region of interest; comparing the pixel value of each pixel point in the set area with the minimum pixel value min and the maximum pixel value max, and setting the pixel value of the pixel point smaller than or equal to the minimum pixel value min as 0; setting the pixel value of the pixel point of which the pixel value is greater than or equal to the maximum pixel value max as 2n-1; setting the pixel value of the pixel point between the minimum pixel value min and the maximum pixel value max to be (value-min)/(max-min) * (2n-1); and generating an annotation file by taking the region of interest as an annotation box. The pixel value of each channel of the region of interest can be expanded to the maximum range through local contrast adjustment, the to-be-labeled object is highlighted to the maximum extent, and the labeling precision is improved; contrast adjusting and marking are carried out at the same time, compared with an existing method of firstly adjusting the contrast and then marking, the contrast adjusting and marking method has the advantages that two steps are shortened into one step, and the marking speed can be increased.

The invention relates to a method and a system for marking reasons of emotion initiation facing multi-user micro-blog, comprising the following steps: automatically acquiring micro-blog data, and constructing a Chinese micro-blog corpus; according to the expression form of emotion triggering reason of micro-blog, a complete set of micro-blog emotion triggering reason tagging method is designed. Designing an image-based tagging system to speed up the tagging of micro-blog emotion triggers; a tagging corpus is obtained by using the tagging system to manually annotate the corpus in the corpus. The invention can mark the reason of emotion contained in the micro-blog from the Chinese micro-blog containing multi-users.

The embodiment of the invention discloses a corpus labeling method and device, electronic equipment and a storage medium. Comprising the steps that corpora to be labeled are obtained, the corpora to be labeled are labeled through a labeling model, an initial labeling result corresponding to the corpora to be labeled is obtained, and the initial labeling result comprises initial labeling corpora corresponding to the corpora to be labeled and labeling information; the labeling information comprises a label value and credibility corresponding to each initial labeling corpus; based on the tag values, the initial tagged corpora are classified to obtain a plurality of classified corpus sets, and the tag values of the initial tagged corpora included in each classified corpus set in the plurality of classified corpus sets are the same; based on the credibility, sorting the initial tagged corpora included in each classified corpus set to obtain a plurality of sorted classified corpus sets; the sorted multiple classified corpus sets are sent to a client side for verification; and receiving the verified tagged corpus returned by the client as a target tagged corpus.

Embodiments of the present invention provide a laser point cloud data recognition method and device, the method comprising: acquiring target laser point cloud data, the target laser point cloud data is based on the laser points in the three-dimensional laser point cloud data in the target three-dimensional scene The following coordinate information is obtained; the target laser point cloud data is input into the trained point cloud deep learning model to obtain the target object recognition result, and the trained point cloud deep learning model is based on the labeling information of the sample target object The obtained model, the annotation information includes the size and position of the sample target object, the annotation information is the information obtained according to the prediction data output by the point cloud depth model, and the prediction data includes the predicted size and position of the sample target object Location. The laser point cloud data recognition method and device provided by the embodiments of the present invention can improve the labeling efficiency, and further improve the recognition efficiency of the target object.