Abstract: Systems and techniques that facilitate hybrid unsupervised and supervised image segmentation are provided. In various embodiments, a system can access a computed tomography (CT) image depicting an anatomical structure. In various aspects, the system can generate, via an unsupervised modeling technique, at least one class probability mask of the anatomical structure based on the CT image. In various instances, the system can generate, via a deep-learning model, an image segmentation based on the CT image and based on the at least one class probability mask.

Abstract: Described are systems and methods for providing a publisher-subscriber architecture for audio and/or video conferencing. Embodiments of the present disclosure can utilize one edge computing resource associated with each client device participating in the audio and/or video conference. One or more client devices can publish streams to the associated edge computing resources. The streams can then be pushed, via a data network, to the edge computing resources associated with the subscribing client devices. The subscribing client devices can access and fetch the streams from the edge computing resources, which can be presented on the client devices as part of a conferencing session experience.

Abstract: A focus adjustment device, that provides a plurality of detection areas in an imaging region, repeatedly generates ranging values for a physical object in the detection areas, and performs focus adjustment for the physical object based on the ranging values, including a processor that has a statistical processing section, a movement state estimating section, a prediction section, and a control section, wherein the statistical processing section subjects ranging values of the plurality of detection areas to statistical processing, the movement state estimating section calculates representative values that contain statistical dispersion based on the statistical processing, and estimates movement state of the physical object based on time-series change in the representative value, the prediction section predicts focal position regarding the physical object based on the movement state that has been estimated, and the control section performs focus adjustment based on the focal position that has been predicted.

Abstract: A method for keypoint matching includes determining a first set of keypoints corresponding to a current environment of the agent. The method further includes determining a second set of keypoints from a pre-built map of the current environment. The method still further includes identifying matching pairs of keypoints from the first set of keypoints and the second set of keypoints based on geometrical similarities between respective keypoints of the first set of keypoints and the second set of keypoints. The method also includes determining a current location of the agent based on the identified matching pairs of keypoints. The method further includes controlling an action of the agent based on the current location.

Abstract: Enhanced methods and systems for the semantic segmentation of images are described. A refined segmentation mask for a specified object visually depicted in a source image is generated based on a coarse and/or raw segmentation mask. The refined segmentation mask is generated via a refinement process applied to the coarse segmentation mask. The refinement process correct at least a portion of both type I and type II errors, as well as refine boundaries of the specified object, associated with the coarse segmentation mask. Thus, the refined segmentation mask provides a more accurate segmentation of the object than the coarse segmentation mask. A segmentation refinement model is employed to generate the refined segmentation mask based on the coarse segmentation mask. That is, the segmentation model is employed to refine the coarse segmentation mask to generate more accurate segmentations of the object. The refinement process is an iterative refinement process carried out via a trained neural network.

Abstract: A system for detecting an object collision includes a processor and a memory in communication with the processor with an objection collision detection module. The objection collision detection module has instructions that, when executed by the processor, cause the processor to obtain a first image and a second image of a scene from a camera sensor mounted to a vehicle, determine a change amount between the first image and the second image, and based on the change amount, determine at least one of that an object is approaching the vehicle or the object has collided with the vehicle. The change amount represents the number of pixels with the same location value that changed intensity values between the first and second images.

Abstract: An automatic threat recognition system and method is disclosed for scanning the x-ray CT image of an article to identify the objects of interest (OOIs) contained within the article, which are otherwise not always quickly apparent or discernable to an individual. The system uses a computer to receive information from two-dimensional (2D) image slices from a reconstructed computed tomography (CT) scan image and to produce a plurality of voxels for each slice of the 2D image. The computer analyzes the voxels to create a likelihood map (LM) representing likelihoods that voxels making up the CT image are associated with a material of interest (MOI). The computer further analyzes the LM to construct neighborhoods of voxels within the LM, and classifies each voxel neighborhood based on its features, thereby decluttering the LM to facilitate the process of connecting voxels of a like MOI together to form segments.

Abstract: Embodiments include a method for ground surface segmentation on sparse Light Detection And Ranging (LiDAR) point clouds comprising: reading a LiDAR point cloud from a LiDAR sensor, the LiDAR point cloud comprising data representing one or more objects in physical surroundings detected by the LiDAR sensor; voxelizing the LiDAR point cloud to produce a three-dimensional representation of each of the one or more objects; constructing a maximum height map from the three-dimensional representation of each of the one or more objects, the maximum height map comprising a two-dimensional mapping of spatial points representing each of the one or more objects; performing minimum filtering on the spatial points of the maximum height map; and classifying each spatial point as a ground point or a non-ground point based on the minimum filtering of each spatial point.

Abstract: In an example embodiment, a process may create a disk for each point of the point cloud, where a size of the disk and the orientation of the disk are respectively based on a computed scale and a non-oriented normal. The process may insert each disk into a search structure that is queried to determine if one or more disks intersect the paths from a given point to each scanner in the point cloud. The process may create an output corresponding to each scanner in the point cloud, where each output includes the number of intersecting disks and the distance from the given point to the scanner. The process may implement a sorting algorithm to assign the given point to a scanner position. An application may utilize the assignment of each point in the point cloud to a scanner position to generate a high-resolution 3D mesh of a scene.

Abstract: Systems, devices, methods, media, and instructions for automated image processing and content curation are described. In one embodiment a server computer system receives a content message from a first content source, and analyzes the content message to determine one or more quality scores and one or more content values associated with the content message. The server computer system analyzes the content message with a plurality of content collections of the database to identify a match between at least one of the one or more content values and a topic associated with at least a first content collection of the one or more content collections and automatically adds the content message to the first content collection based at least in part on the match. In various embodiments, different content values, image processing operations, and content selection operations are used to curate content collections.

Abstract: Aligning multiple 3D images of an object can be difficult when the representative datasets (images) are large. An exemplary aspect of this technology teaches a technique to subdivide the images and use the alignments between the subdivided images to determine the alignment between the complete datasets.

Abstract: A system and a method for automatically clustering sandbox analysis reports of similar malware samples. An automated malware analysis process includes receiving from a sandbox server the sandbox analysis reports of the similar malware samples at an application programming interface (API) of the clustering server, clustering similar Uniform Resource Locators (URLs) together and clustering the sandbox analysis reports of events in sandbox reports clusters (1-n) based on the URL clustering, static properties of the malware samples and dynamic properties of the malware samples.

Abstract: Systems and methods for identifying content for an input query are presented. A mapping model is trained to map elements of an input query embedding vector for a received query into one or more elements of a destination embedding vector. In response to receiving an input query, an input query embedding vector is generated that projects into an input query embedding space. The input query embedding vector is processed by the mapping model to map the input query embedding vector into one or more elements of a destination embedding vector in a destination embedding space, resulting in a partial destination embedding vector. Items of a corpus of content are projected into the destination embedding space and the partial destination embedding vector is also projected into the destination embedding space. A similarity measure determines the most-similar items to the partial destination embedding vector and at least some of the most-similar items are returned in response to the input query.

Abstract: A computer implemented method for object detection includes the following steps carried out by computer hardware components: acquiring an image; determining a pixel of the image as a base pixel; determining coordinates of a plurality of sets of target pixels, each set of target pixels including a plurality of pixels in a respective pre-determined relationship to the base pixel; for each of the sets of target pixels, determining information representing values of the pixels in the respective set of target pixels; and determining whether a pre-determined object is shown in the image based on the determined information.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In one example process a set of words including a grammatical error is received. The process can generate, using a neural network based on the set of words including the grammatical error and a reference set of words, a transformed set of words and further determine, based on the set of words including the grammatical error and the reference set of words, a reconstructed reference set of words. The process can also determine, based on a comparison of the transformed set of words and the reconstructed reference set of words, whether the transformed set of words is grammatically correct and provide an indication of whether the transformed set of words is grammatically correct to the neural network.

Abstract: Users often desire to capture certain images from an application. Existing methods of capturing images can result in low-resolution images due to limitations of the display device providing the images. This disclosure provides a method of capturing higher resolution images from source images. Techniques are also disclosed to reduce the storage size associated with the higher resolution images. Through capturing low-resolution versions of the same source images, image effects can be captured and applied to the higher resolution images where those image effects may be altered or missing. Frequency spectrum combination can be used to combine the low-resolution image data and the higher resolution image data. The higher resolution images can be processed using a segmentation scheme, such as tiling, without reducing or limiting the image effects.

Abstract: An image processing method includes receiving an image frame, detecting a face region of a user in the image frame, aligning a plurality of preset feature points in a plurality of feature portions included in the face region, performing a first check on a result of the aligning based on a first region corresponding to a combination of the feature portions, performing a second check on the result of the aligning based on a second region corresponding to an individual feature portion of the feature portions, redetecting a face region based on a determination of a failure in passing at least one of the first check or the second check, and outputting information on the face region based on a determination of a success in passing the first check and the second check.

Abstract: The present invention comprises the capture and display of arthropod, human and arthropod-based metadata, which is capable of tracking and displaying the metadata, which is time and location-based, in order to show migration paths of arthropods and/or the diseases they have the potential to carry. This real-time view can help predict future arthropod and disease based on various scenarios such as, but not limited to: increased exposure based on the following: a user's geo-location, date and/or time of year, carrier type, etc. These variables can then assist with the education, awareness and potential prevention of disease.

Abstract: The present invention relates to a scene descriptor generation system, including an image sensor including a plurality of pixels, successively supplying S binary image planes, each including a single binary value per pixel; and a processing circuit configured to, for each binary image plane supplied by the sensor (100), to implement the successive steps of: a) calculating N convolutions of the binary image plane with respectively N distinct convolution kernels, to provide N convolved images; b) generating at least one meta-image from the N convolved images supplied at step a); and c) for each of the N convolved images supplied at step a), generating a binary descriptor from said convolved image and from said at least one meta-image.

Abstract: Various embodiments of systems and methods for image matching and retrieval process are disclosed.

Abstract: A method includes receiving a first operation instruction; responsive to receiving the first operation instruction, determining whether one or more first images from a set of images comprise a first common feature; and responsive to determining that the one or more first images from the set of images comprise the first common feature, displaying the one or more first.

Abstract: Systems and methods are provided for automatically inspecting photovoltaic (PV) installations. The system utilizes drones flying preprogrammed routes to conduct aerial thermography of PV modules for inspection (i.e., gathering IR images of the PV modules). The system conducts analytics on the gathered images to determine defects in the PV modules of the PV installation, classifies detected defects, issues reports automatically, and retains technical data of every module included in the database of a PV installation for trend analysis and preventative and predictive maintenance analysis.

Abstract: This disclosure describes systems and methods for matching user provided images that include representations of items with sellers of those items. A management service, as described herein, may provide a web site where users can post images, view images, share images, correspond with other users, etc. The management service may identify items represented in the images and determine one or more sellers that offer those items for sale. When another user requests to view the user provided image, the image, seller information identifying the seller determined to sell the item represented in the image, and/or a purchase control that may be selected by a user to initiate a purchase with the seller is presented.

Abstract: A system and method for removing contextually identical multimedia content elements. The method includes analyzing a plurality of multimedia content elements to identify at least two multimedia content elements of the plurality of multimedia content elements that are contextually identical; selecting, from among the at least two contextually identical multimedia content elements, at least one optimal multimedia content element; and removing, from a storage, all multimedia content elements of the group of contextually identical multimedia content elements other than the at least one optimal multimedia content element.

Abstract: A method for grasping texture-less metal parts based on BOLD image matching comprises: obtaining a real image and CAD template images by photographing, extracting a foreground part of the input part image, calculating a covariance matrix of a foreground image, establishing the direction of a temporary coordinate system, and setting directions of line segments to point to a first or second quadrant of the temporary coordinate system; constructing a descriptor of each line segment according to an angle relation between the line segment and k nearest line segments, and matching the descriptors of different line segments in the real image and the CAD template images to obtain line segment pairs; and recognizing a processed pose through a PNL algorithm to obtain a pose of a real texture-less metal part, and then inputting the pose of the real texture-less metal part to a mechanical arm to grasp the part.

Abstract: A computer-based method for simulating a human-like decision in an environmental context, comprising: capturing environmental data with at least one sensor, realising a computer based method for realising a bi-directional compression of high dimensional data by compressing the data into a lower-dimensional map, if environmental data are captured during a learning phase of a computer-based model, evaluate the map of compressed data by determining the quality of the map by how well it separates data with different properties, the captured data corresponding to known pre-recorded data that have been pre-evaluated, if environmental data are captured after the learning phase, add new point to the compressed data and generate a signal indicating which human-like decision to use to correspond to the state of the operator.

Abstract: A computer-implemented method for generating a feature descriptor for a location in an image for use in performing descriptor matching in analysing the image, the method comprising determining a set of samples characterising a location in an image by sampling scale-space data representative of the image, the scale-space data comprising data representative of the image at a plurality of length scales; and generating a feature descriptor in dependence on the determined set of samples.

Abstract: A system for tracking airborne organisms includes an imager, a backlight source (such as a retroreflective surface) in view of the imager, and a processor configured to analyze one or more images captured by the processor to identify a biological property of an organism.

Abstract: Systems, methods and apparatus of vehicle security operations during parking. For example, a vehicle includes: a proximity sensor configured to detect presence of an object approaching the vehicle when the vehicle is in a parking state; at least one camera configured to monitor surroundings of the vehicle when the vehicle is in the parking state; and an artificial neural network configured to extract identification information of the object from images generated by the camera and determine a security classification of the presence of the object. The identification information is stored in the vehicle and/or transmitted to a server or a mobile device, in response to the security classification being in a predefined category.

Abstract: A passage determination method includes: determining a target space region corresponding to a vehicle in a driving direction based on a real-time road image of the driving direction in a real scene; determining an actual size of the target space region; and determining whether the vehicle is able to pass through the target space region according to a magnitude relationship of the actual size and a safe size of the vehicle.

Abstract: An approach for an iris liveness detection is provided. A plurality of image pairs is acquired using one or more image sensors of a mobile device. A particular image pair is selected from the plurality of image pairs, and a hyperspectral image is generated for the particular image pair. Based on, at least in part, the hyperspectral image, a particular feature vector for the eye-iris region depicted in the particular image pair is generated, and one or more trained model feature vectors generated for facial features of a particular user of the device are retrieved. Based on, at least in part, the particular feature vector and the one or more trained model feature vectors, a distance metric is determined and compared with a threshold. If the distance metric exceeds the threshold, then a first message indicating that the plurality of image pairs fails to depict the particular user is generated.

Abstract: A danger ranking training method comprising training a first deep neural network for generic object recognition within generic images, training a second deep neural network for specific object recognition within images of a specific application, training a third deep neural network for specific scene flow prediction within image sequences of the application, training a fourth deep neural network for potential danger areas localization within images or image sequences of the application using at least one human trained danger tagging method, training a fifth deep neural network for non-visible specific object anticipation and/or visible specific object prediction within image or image sequences of the application, and determining at least one danger pixel within an image or an image sequence of the application using an end-to-end deep neural network as a sequence of transfer learning of the five deep neural networks followed by one or several end-to-end top layers.

Abstract: A computing device of a first vehicle may receive a first image and a second image of a second vehicle having flashing light signals. The computing device may determine, in the first image and the second image, an image region that bounds the second vehicle such that the image region substantially encompasses the second vehicle. The computing device may determine a polar grid that partitions the image region in the first image and the second image into polar bins, and identify portions of image data exhibiting a change in color and a change in brightness between the first image and the second image. The computing device may determine a type of the flashing light signals and a type of the second vehicle; and accordingly provide instructions to control the first vehicle.

Abstract: An image search device receives an image search request and selects, from a plurality of images, a first group of images that match search conditions based on the search request. The image search device further selects, from the first group of images, a second group of images based on a score calculated by a predetermined evaluation formula, and outputs information about the second group of images as a search result. The predetermined evaluation criteria may be set so as to increase the score for an image in the first group of images based on positive engagement by many people on the predetermined internet service and based on the extent that the image is less similar to other images in the first group of images.

Abstract: A memory stores a first log and a collection of queries. A processor accesses a second log that includes a first, second, third and fourth historical query, determines that the first and second historical queries belong to a first similarity group and the third and fourth historical queries belong to a second similarity group, and ranks the queries within groups based on static and dynamic features. The processor further determines the third historical query is frequently submitted and copies the highest ranked similar queries into the collection. The processor receives a first database query, stores it in both logs, determines it is not similar to queries stored in memory, further determines that it belongs to the first similarity group, copies the highest ranked similar queries to the first log, and returns a report containing the similar queries. The processor further receives and executes a second query selected from the report.

Abstract: The present technology can analyze a collection of images to generate a high-dimension first representation of each image in the collection of images, and then reduce the dimensionality of the high-dimension first representation of each image by identifying significant features from the high-dimension first representation to yield a reduced-dimension second representation for each image in the collection of images. The present technology can then compare the second representation of any image in the collection of images to the second representation of any other image in the collection of images, wherein images having identical second representation are duplicates.

Abstract: The image processing apparatus has: an image acquisition unit configured to acquire a captured image of an image capturing area in which an object is located; a determination unit configured to determine parameters to be used for image processing to improve resolution of an image of the object, by learning using a dataset of images of the object, wherein the dataset is generated based on object information indicating a degree of importance of the object; and a processing unit configured to perform the image processing to improve the resolution of an image of the object included in the acquired captured image, using the parameters determined by the determination unit.

Abstract: A method for creating content includes making a corpus of images available, adding a plurality of tag data to each of the images, receiving a query, extracting a trigger from the query, identifying a set of the images in the corpus, wherein identified images have tag data matching the extracted trigger, and creating a video comprising the identified images.

Abstract: An image search device stores images and tag information about the images, receives an image, extracts, from the image, feature values for search queries a similar image search unit which performs a similar-image search to obtain similar images and grouping information, divides the queries into groups in accordance with the obtained grouping information, calculates the levels of importance of the groups, sorts the search results in accordance with the levels of importance of the groups, and outputs information about the search results for each group.

Abstract: Systems and methods are presented for efficient and effective computation of LSH hashing functions for use in various environments, including nearest-neighbor search. A corpus of items is maintained, with each item of the corpus being associated with information identifying one of a plurality of LSH partitions to which the item has been mapped or hashed. In response to an input item, the input item is projected one or more times into an orthoplex inscribed within a multi-dimensional space. For each mapping, a determination is made as to the nearest vertex of the orthoplex to the projected input item in view of determined offsets that provide for balanced distribution, and the LSH partitions associated with the nearest vertex are aggregated and associated with the input item as the input item's LSH partition.

Abstract: A method for detecting objects of a specified class in an aerial image of an area, including manipulating first and second features related to each element of the image according to image metadata and pre-configured rules, selecting a first group of elements complying with a first condition related to the first feature, the first condition based on the image metadata and the pre-configured rules, selecting a second group of elements complying with a second condition related to the second feature, the second condition based on the image metadata and the pre-configured rules, and creating a representation of the area from elements common to the first group and the second group, the representation including elements belonging to the objects of the specified class.

Abstract: An image processing device 10 includes: a feature extraction unit 11 which obtains features in each of scaled samples of the region of interest in a probe image; a saliency generation unit 12 which computes the probabilities of the pixels in the scaled samples that contribute to the score or the label of the object of interest in the region; a dropout processing unit 13 which removes the features from the scaled samples which are not essential for the computing the score or the label of the object, using the computed probabilities.

Abstract: A lane line processing method and device are provided. The method can includes obtaining distances between lane line points in a first image, determining direction densities of the individual lane line points by using the distances between the lane line points, dividing lane line points into groups corresponding to lane lines by using the direction densities of the individual lane line points, and obtaining representation information of the lane lines corresponding to the groups by using the lane line points in the groups. By using the direction densities of the individual lane line points, lane line points in an image are divided into groups, the obtained groups are then more accurate, and thus the lane lines obtained by fitting are accurate, are less susceptible to image quality issues, and have high robustness.

Abstract: The present disclosure provides a method and an apparatus for object re-identification, capable of solving the problem in the related art associated with inefficiency and low accuracy of object re-identification based on multiple frames of images.

Abstract: A method is disclosed of discriminating detected objects in an area with a vision apparatus. The method includes generating image data of a portion of the area using an imaging device of the object detection device, and processing the image data to classify the image data as an imaged scene type selected from a plurality of scene types stored as scene type data in the memory. The method further includes processing the image data using the object identification data to generate object detection data for each object of the plurality of objects located in the portion of the area, each object detection data having a corresponding scene type of the plurality of scene types obtained from the object identification data, and generating user a sensible output only for the object detection data having a corresponding scene type that is the same as the imaged scene type.

Abstract: An apparatus that identifies a scene type includes at least one processor and a memory. The memory is operatively coupled to the at least one processor and is configured to store instructions executable by the processor. Upon execution of the instructions, the processor is caused to identify a scene type of content that includes video and audio based on a feature amount of the audio in the content.

Abstract: Systems, devices, methods, media, and instructions for automated image processing and content curation are described. In one embodiment a server computer system receives a content message from a first content source, and analyzes the content message to determine one or more quality scores and one or more content values associated with the content message. The server computer system analyzes the content message with a plurality of content collections of the database to identify a match between at least one of the one or more content values and a topic associated with at least a first content collection of the one or more content collections and automatically adds the content message to the first content collection based at least in part on the match. In various embodiments, different content values, image processing operations, and content selection operations are used to curate content collections.

Abstract: In an exemplary process for providing content in an augmented reality environment, image data correspond to a physical environment are obtained. Based on the image data, predefined entities of the plurality of predefined entities in the physical environment are identified using classifiers corresponding to predefined entities. Based on the one or more of the identified predefined entities, a geometric layout of the physical environment is determined. Based on the geometric layout, an area corresponding to a particular entity is determined. The particular entity corresponds to one or more identified predefined entities. Based on the area corresponding to the particular entity, the particular entity in the physical environment is identified using classifiers corresponding to the determined area. Based on the identified particular entity, a type of the physical environment is determined.

Abstract: A method of complementary clustering of a vast population of objects is disclosed. The method aims at maximizing a global measure of object affinity within naturally-formed clusters. A first clustering procedure produces primary centroids of clusters of objects and a second clustering procedure produces secondary clusters of the primary centroids and corresponding secondary centroids. Refined clusters of the population of objects are formed based on object proximity to the secondary centroids. The first clustering procedure is preferably based on a variation of the K-means method, and the second clustering procedure is preferably based on the Density-Based Spatial Clustering of Applications with Noise (DBSCAN). An apparatus implementing the method is devised to facilitate conflict-free parallel processing.

Abstract: Systems and methods for digitized document image text contouring are provided. One or more memories may be coupled to one or more processors, the one or more memories including instructions operable to be executed by the one or more processors. The one or more processors may be configured to receive a digitized document image. The one or more processors may be configured to preprocess the digitized document image to generate a plurality of contours. The one or more processors may be configured to adjust a plurality of bounding boxes of the plurality of contours; analyze the adjusted plurality of bounding boxes; create one or more clips based on the analysis; dynamically threshold the one or more clips; perform optical character recognition of the one or more clips; and receive output responsive to the optical character recognition.