435 results about "Image labeling" patented technology

An improved image tagging system that tags images in a mobile handheld communications device, based on the user's contacts list stored therein. The image tagging system compares a face in a given picture (that the user has just taken using the built-in camera) to faces in the user's contacts list images. A matching contact is found in the contacts list. The picture is then tagged with information from the matching contact. The picture is tagged by inserting identification information into a header of the picture.

A global tag for an image is received identifies one or more objects in the image, and a region-specific tag for the image identifies one or more objects in a region of the image. The global tag and the region-specific tag are stored with the image. Displayed, along with the image, is an identifier for each of the one or more objects identified in the global tag, and an identifier for each of the one or more objects identified in the region-specific tag. Different users are able to maintain different names for the same person, allowing the same tag of the image to be used as the basis for displaying the image with different names for the different users. Additionally, the tags can be used as a basis for generating a credits list of people that are included in a compilation of images.

Identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. A technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. One example embodiment trains a convolutional network including multiple layers of filters. The filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. The filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. The technique can be applied to images of neural circuitry or electron microscopy, for example. The same technique can also be applied to correction of photographs or videos.

A system for and method of enhancing web page delivery and visitor tracking. The invention provides the ability to control redirection of Web traffic of humans. It can identify visitors to a Web page, track their movements, log critical information, and analyze the Web traffic in order to judge the success in driving quality traffic to some known goal on a Web site, such as a sale. The system tracks visitors via use of redirection with an image tag. The invention can also use unique image marker references to prevent browsers from accessing cached images.

An apparatus for labelling images comprises a segmentation processor (103) which segments an image into image segments. A segment label processor (105) assigns segment labels to the image segments and a relation processor (107) determines segment relations for the image segments. A CRP model processor (109) generates a Constraint Reasoning Problem model which has variables corresponding to the image segments and constraints reflecting the image segment relations. Each variable of the model has a domain comprising image segment labels assigned to an image segment of the variable. A CRP processor (111) then generates image labelling for the image by solving the Constraint Reasoning Problem model. The invention may allow improved automated labelling of images.

A method of registering a print between print stages in a printing system is provided. The method comprises: printing a first part of a digital image on a substrate with a first print engine, wherein the first part of the digital image includes a plurality of reference features; scanning image content on the substrate as it is readied for the second print engine; determining the actual positions of the reference features; using the actual positions of the reference features to determine desired adjustments for the second part of the digital image; applying electronic registration according to the desired adjustments to modify image content of the second print to be aligned with content already on the substrate; and printing a second part of the digital image on the substrate.

Object recognition techniques are disclosed that provide both accuracy and speed. One embodiment of the present invention is an identification system. The system is capable of locating objects in images by searching for local features of an object. The system can operate in real-time. The system is trained from a set of images of an object or objects. The system computes interest points in the training images, and then extracts local image features (tokens) around these interest points. The set of tokens from the training images is then used to build a hierarchical model structure. During identification / detection, the system, computes interest points from incoming target images. The system matches tokens around these interest points with the tokens in the hierarchical model. Each successfully matched image token votes for an object hypothesis at a certain scale, location, and orientation in the target image. Object hypotheses that receive insufficient votes are rejected.

A system for monitoring an image tagging system comprises a processor and a memory. The processor is configured to provide a user display for a type of monitoring of an image tagging system. The user display includes image tagging monitoring for: a) in the event that a computer vision match is found, a computer vision match and an associated computer vision tag and, b) in the event that no computer vision match is found, a human vision match and an associated human vision tag. A memory coupled to the processor and configured to provide the processor with instructions.

The embodiment of the invention provides a method for labeling medical images. The method includes the steps of dividing an unlabeled medical image set into at least two unlabeled medical image subsets, distributing the unlabeled medical image subsets to at least two labeling terminals so that each labeling terminal can label the medical images in the corresponding distributed unlabeled medical image subset, and receiving labeling information unloaded by each labeling terminal. Accordingly, the embodiment of the invention further provides a device and system for labeling the medical images. Through the method, device and system, the coordinative medical image labeling solution can be obtained, and the labeling efficiency is improved.

The embodiment of the invention discloses a method, device and system for automatically labeling a target object in an image, and the method comprises the steps: obtaining an image training sample which comprises a plurality of images, each image is obtained through the shooting of the same target object, and the adjacent images have the same environmental feature points; Taking one image as a reference image, determining a reference coordinate system, and creating a three-dimensional space model based on the reference three-dimensional coordinate system; When the three-dimensional space modelis moved to the position where a target object is located in the reference image, determining position information of the target object in the reference three-dimensional coordinate system; And respectively mapping the three-dimensional space model to the image plane of each image according to the respective corresponding camera attitude information determined by the environmental feature pointsin each image. According to the embodiment of the invention, automatic image labeling can be carried out more accurately and effectively, and the universality of the method is improved.

A method of processing image data includes providing an image capturing apparatus. A host device for receiving image data generated by the image capturing apparatus is provided. An image to be captured with the image capturing apparatus is targeted. First image data representing the targeted image is generated using the image capturing apparatus. The first image data has a first resolution size. A first tag to the first image data is attached for post processing in the host device. The first tag instructs the host device to convert the first image data from the first resolution size to a second resolution size. The apparatus is programmed to automatically attach the first tag to the first image data once the first image data is generated.

A method for learning visual attribute labels for images includes, from textual comments associated with a corpus of images, identifying a set of candidate textual labels that are predictive of aesthetic scores associated with images in the corpus. The candidate labels in the set are clustered into a plurality of visual attribute clusters based on similarity and each of the clusters assigned a visual attribute label. For each of the visual attribute labels, a classifier is trained using visual representations of images in the corpus and respective visual attribute labels. The visual attribute labels are evaluated, based on performance of the trained classifier. A subset of the visual attribute labels is retained, based on the evaluation. The visual attribute labels can be used in processes such as image retrieval, image labeling, and the like.

The invention discloses an intelligent image annotation method based on YOLOv3 depth learning network, belonging to the technical field of image retrieval. Firstly, the image data set to be labeled bythe image recognition item is divided into M groups, one group is manually labelled, and it is used to train YOLOv3 deep learning network, then, the obtained network model is automatically labeled with another set of unlabeled image data, and added into the training set of YOLOv3 to continue training, and the labeling and training process is repeated until all sets of image data are labeled successfully, and the updated model of the whole data set corresponding to the image recognition item is obtained. The invention realizes intelligent labeling of picture data sets of image recognition items, not only effectively reduces the complex workload of image labeling, but also gradually improves the performance of YOLOv3 model detection and recognition target through several cyclic iterative training.

The invention relates to a slice image processing method and device, computer equipment and a storage medium. The method comprises the following steps: segmenting a slice image into a plurality of cell images according to acquired image segmentation parameters; inputting the cell image into the trained convolutional neural network model; obtaining classification probability data, and according tothe classification probability data, determining a classification category corresponding to the cell image, then counting the number of cell images of an abnormal category in the slice image, calculating the proportion data of the counted number of cell images in the total number of cell images, and when the proportion data is greater than a preset proportion threshold, marking the slice image asan abnormal slice image. On one hand, detailed analysis is carried out by segmenting the slice image. The overall analysis accuracy of the slice image is improved. On the other hand, classification accuracy of a single cell image is improved by adopting an artificial intelligence convolutional neural network, so that the analysis accuracy of the slice image is improved.

The invention discloses a character image identifying and labeling method and system. The method or system is realized through the following steps of: firstly obtaining an original character image ora database comprising a plurality of character image labeling samples; secondly, establishing a Laplace additional information-against neural network model and generating an artificially synthesized character image according to the original character image and the Laplace additional information-against neural network model; and finally classifying the artificially synthesized character image according to an image type so as to obtain a feature image, matching the database comprising multiple character image labeling samples with the feature image, and adding labelling information of a character image labeling sample which is mostly matched with the feature image onto the feature image so as to realize automatic classification and labeling of the original character image and then improve the character image identifying and labeling efficiency.

A system for tagging an image comprises a processor and a memory. The processor is configured to analyze an image associated with an image query using one or more computer vision analysis types to determine zero or more computer vision matches. Each computer vision match has one or more associated computer vision tags. In the event that it is determined that there are zero computer vision matches, the processor is further configured to analyze the image associated with the image query using a human vision analysis system to determine zero or more human vision matches. Each human vision match has one or more associated human vision tags. A memory coupled to the processor and configured to provide the processor with instructions.

The invention discloses a medical image labeling method and device for deep learning, and the method comprises the following steps: carrying out the preprocessing of an inputted image diagnosis report, electronic medical record information and a medical image, and generating data comprising the medical image and the extracted related diagnosis information, wherein the medical image is pre-annotated based on deep learning, the image is segmented based on an image semantic segmentation technology to obtain a boundary range of each lesion area, pixel-level segmentation annotation is performed onthe image, and classification annotation is performed on a disease type to which the image belongs based on an image classification technology in combination with image related diagnosis information;and displaying the pre-annotated image and the related diagnosis information through an interface to receive an interactive instruction for a doctor to finely adjust a pre-annotated image result, andexporting an annotation result. According to the method, the labeling efficiency is improved, and the labeling precision is ensured.

The invention discloses a semi-automatic image tagging method based on semantics and content. The tagging method comprises the following steps: a feedback image retrieval system based on semantics and content is established; a model of the system is expressed as (F, Q, R (fq, fd)) with a triple; a user submits an query based on keywords; the system converts a keyword sequence into a group of semantic characteristics vector Omega i', uses an arrangement function R (fq, fd)) for doing an query based on semantics for all the images in a database, and returns the query result to the user according to the arrangement; the user chooses relatively satisfactory images from the query result and feeds back to the system; the system again uses the arrangement function R (fq, fd)) for doing the query based on semantics or content or integrating the semantics and the content for all the images in a database, finds out all the images which are similar to the images that the user is satisfied with, and simultaneously adjusts the weight of the tagging information that each image corresponds to. The invention has the advantages of high efficiency, high accuracy, and friendly interactive way.

Multi-scale processing may be used to reduce the memory and computational requirements of optimization algorithms for image labeling, for example, for object segmentation, 3D reconstruction, stereo correspondence, optical flow and other applications. For example, in order to label a large image (or 3D volume) a multi-scale process first solves the problem at a low resolution, obtaining a coarse labeling of an original high resolution problem. This labeling is refined by solving another optimization on a subset of the image elements. In examples, an energy function for a coarse level version of an input image is formed directly from an energy function of the input image. In examples, the subset of image elements may be selected using a measure of confidence in the labeling.

The invention discloses an image grey level histogram-based foggy day detection method, which mainly comprises: a first step of performing initialization to obtain a grey level histogram of an image; a second step of primarily detecting whether the image is marked to indicate a foggy day or a fogless day; a third step of performing processing again if the image is marked to indicate the fogless day, and marking the image to indicate the foggy day when a certain condition is met; a fourth step of further performing detection if the image is marked to indicate the foggy day, and marking the image to indicate the fogless day when the certain condition is met; and a fifth step of performing the detection for the third time if the image is marked to indicate the foggy day, marking the image to indicate a densely-foggy day when the certain condition is met, otherwise, marking the image to indicate a thinly-foggy day. In the method, the grey level histogram of the image is utilized to detect weather for the first time, and three levels which are the fogless day, the thinly-foggy day and the densely-foggy day respectively are detected by utilizing corresponding relationships between the number of pixels and grey level values in the grey level histogram and a series of threshold values. Compared with other foggy day detection methods, the method has the advantages of low cost, easy popularization, high processing speed, wide application range, high accuracy and ideal effect.

Identifying objects in images is a difficult problem, particularly in cases an original image is noisy or has areas narrow in color or grayscale gradient. A technique employing a convolutional network has been identified to identify objects in such images in an automated and rapid manner. One example embodiment trains a convolutional network including multiple layers of filters. The filters are trained by learning and are arranged in successive layers and produce images having at least a same resolution as an original image. The filters are trained as a function of the original image or a desired image labeling; the image labels of objects identified in the original image are reported and may be used for segmentation. The technique can be applied to images of neural circuitry or electron microscopy, for example. The same technique can also be applied to correction of photographs or videos.

A system and method for estimating a set of landmarks for a large image ensemble employs only a small number of manually labeled images from the ensemble and avoids labor-intensive and error-prone object detection, tracking and alignment learning task limitations associated with manual image labeling techniques. A semi-supervised least squares congealing approach is employed to minimize an objective function defined on both labeled and unlabeled images. A shape model is learned on-line to constrain the landmark configuration. A partitioning strategy allows coarse-to-fine landmark estimation.

A printing system is provided for enabling a consistent color job re-print comprising a plurality of image marking engines. The system includes printing a master test image document printed by at least a first image marking engine. The master test image can have a data glyph and a plurality of reference patches. The data glyph can record distinct job attributes and the plurality of reference patches can record color parameters of a desired output. The system further provides a scanner for scanning the master test image document. The data glyph can provide the job attributes of the master test image document and can generate set-up data therefrom. The reference patches can provide the color parameters of the master test image document and can generate set-up data therefrom.

The invention discloses a gender recognition method based on a convolution neural network, and the method comprises the steps: obtaining a training sample set, and setting gender labels for all human face images; carrying out the preprocessing of each human face image sample, and obtaining Gabor features in different directions at different scales; obtaining a plurality of Gabor feature images, and converting the plurality of Gabor feature images into a one-dimensional feature vector; carrying out the reduction of dimensions, and converting the one-dimensional feature vector into a feature matrix which is matched with the size of an input layer of the convolution neural network; obtaining the convolution neural network according to the feature matrix of the human face image sample and the gender labels through training; extracting the corresponding feature matrix of a to-be-recognized human face image through the same method, inputting the feature matrix into the trained convolution neural network, and obtaining a gender recognition result. The method employs the Gabor features and combines the convolution neural network for the gender recognition, improves the robustness of illumination changes, and improves the recognition rate of gender.

The invention provides an image labeling method and device and a storage medium, and the method comprises the steps: receiving a to-be-detected image input by a user, the to-be-detected image comprising a target object located in a target scene; and adopting an object detection model to mark a target object in the to-be-detected image, the object detection model being obtained by training according to the to-be-trained image set, the to-be-trained image set comprising a plurality of to-be-trained images synthesized by an image comprising the same scene type as the target scene and an image comprising the target object. According to the invention, the image in the to-be-trained image set used by the object detection model adopted by the image labeling is synthesized by the image containing the target object and the scene of which the scene type is the same as that of the target scene; the image does not need to be labeled in advance, the efficiency of obtaining the object detection model is improved, and then the image labeling efficiency is improved.

The invention relates to an evaluation system and method for the efficacy of synchro-neoadjuvant chemoradiotherapy before the rectal cancer surgery. The evaluation system comprises an image acquisition unit used for acquiring pathological biopsy slice scan images and neoadjuvant chemoradiotherapy treatment pre-MRI images of newly diagnosed locally advanced rectal cancer patients, and classifying the rectal cancer patients into a training set, a check set and a test set, to serve as input image data, an image labeling unit used for respectively labeling the pathological biopsy slice scan imagesand the MRI images of the training set, the check set and the test set, a convolutional neural network constructing unit used for constructing a first convolutional neural network model, and a convolutional neural network model training unit used for obtaining a second convolutional neural network model for evaluating the efficacy of synchro-neoadjuvant chemoradiotherapy before the rectal cancersurgery. The evaluation system for the efficacy of synchro-neoadjuvant chemoradiotherapy before the rectal cancer surgery has multiple advantages such as being high in accuracy, short in time consumption, long in working duration, objective and three-dimensional.

The invention discloses an image base establishment method and system, an image base and an image classification method. The image base establishment method comprises the steps of obtaining an image to be labeled; determining an initial label of the image according to an image recognition algorithm; obtaining a user-defined label of the image, wherein the user-defined label is input text of a user; determining a target label of the image according to the initial label and the user-defined label; storing the image and the target label, and forming a training sample. According to the image base establishment method, firstly, the initial label of the image is determined according to the image recognition algorithm; then, the image is labeled according to the initial label and the user-defined label labeled by the user, so that the professionalism of image labeling is improved; meanwhile, and with worker's opinions for reference, the labeling quality and the purity of the training sample are improved.

The invention provides a segmentation method and device, a storage medium and terminal equipment for a video image. The segmentation method of the video image comprises the following steps that the heatmap labeling information and the frame image of the previous frame of image in the video are acquired; the heatmap labeling information and the frame image of the previous frame of image are input into a neural network model, and the heatmap labeling information of the frame of image is output. The neural network model is used for representing the heatmap labeling information of the previous frame of image, the frame image, and the correlation between the frame image and the thermal graph labeling information of the frame image. The frame image is segmented according to the thermal image labeling information of the frame image. According to the invention, the complexity of the neural network model is reduced, and the neural network model is suitable for being applied to the mobile terminal.

A method of color management for image marking devices utilizes an automated spot color editor having a control module accessing a graphical user interface. The method includes receiving image data input, in either device-dependent color space or device-independent color space, for a marking job. It is determined whether spot colors are present within the image data input and whether the CMYK values for each of the spot colors present within the image data input are included in the marking device spot color dictionary. Operational parameters for the automated spot color editor are initialized, with operational parameters including the desired performance criteria to be minimized by the automated spot color editor through selection of one or more matrices from a plurality of possible gain matrices to identify new CMYK values. The quality level of the new CMYK values is assessed and new CMYK values are transmitted to image printing device(s).