Abstract: A system for identifying products in a retail store is provided. The system may include a processor configured to: receive an image depicting a store shelf and a product displayed thereon; select a product model subset based on a characteristic of the store shelf determined based on analysis of the image; determine whether the product model subset is applicable to the product; and when the product model subset is applicable to the product, analyze a representation of the product depicted in the image using the product model subset, and identify the product based on the analysis of the representation of the product.

Abstract: Embodiments relate to a multi-mode demosaicing circuit able to receive and demosaic image data in a different raw image formats, such as Bayer raw image format and Quad Bayer raw image format. The multi-mode demosaicing circuit demosaics Quad Bayer image data by interpolating a green channel of the image data along each of a plurality of directions, generating a gradient of the image data along each of the plurality of directions, modifying the interpolated green channels based on respective gradients to generate full-resolution green channel image data, which is combined with red and blue image data to generate the demosaiced image data. Interpolation is performed for non-green pixels based on neighboring green pixels along a specified direction, modified by a residual value based upon values of one or more nearby same-color pixels and a correlation between values of the same-color pixels and neighboring green pixels.

Abstract: A processor-implemented liveness test method includes detecting a face region in a query image, the query image including a test object for a liveness test, determining a liveness test condition to be applied to the test object among at least one liveness test condition for at least one registered user registered in a registration database, determining at least one test region in the query image based on the detected face region and the determined liveness test condition, obtaining feature data of the test object from image data of the determined at least one test region using a neural network-based feature extractor, and determining a result of the liveness test based on the obtained feature data and registered feature data registered in the registration database and corresponding to the determined liveness test condition.

Abstract: A system and method for identifying observations using images of a construction site. A method includes identifying observations in visual media content by identifying items shown in the visual media content, wherein each identified observation corresponds to at least one item of the identified items, wherein each observation corresponds to a type; determining a likelihood impact parameter for each of the observations by applying likelihood rules based on a classification of each of the corresponding at least one item and a ratio of historical incidents of the same type as the observation to a total number of projects in a time period; determining a severity impact parameter for each of the observations by applying severity rules based on an average cost of the historical incidents of the same type as the observation; and determining at least one risk factor for each observation based on its likelihood and severity impact parameters.

Abstract: Disclosed are a method and device for recognizing a video. One specific embodiment of the method comprises: obtaining a video to be identified; inputting said video to a pre-trained local and global representation propagation LGD model to obtain the category of said video, wherein the LGD model learns a spatial-temporal representation in said video based on diffusion between local and global representations. According to this embodiment, the spatial-temporal representation in the video is learned based on diffusion between the local and global representations.

Abstract: Embodiments are disclosed for a method. The method includes receiving an annotation set for a machine learning model. The annotation set includes multiple data points relevant to a task for the machine learning model. The method also includes determining total weights corresponding to the data points. The total weights are determined based on multiple ordering constraints indicating multiple data classes and corresponding weights. The corresponding weights represent a relative priority of the data classes with respect to each other. The method further includes generating an ordered annotation set from the annotation set. The ordered annotation set includes the data points in a sequence based on the determined total weights.

Abstract: A method can include receiving (1) images of at least one subject and (2) at least one total mass value for the at least one subject. The method can further include executing a first machine learning model to identify joints of the at least one subject. The method can further include executing a second machine learning model to determine limbs of the at least one subject based on the joints and the images. The method can further include generating three-dimensional (3D) representations of a skeleton based on the joints and the limbs. The method can further include determining a torque value for each limb, based on at least one of a mass value and a linear acceleration value, or a torque inertia and an angular acceleration value. The method can further include generating a risk assessment report based on at least one torque value being above a predetermined threshold.

Abstract: A method of in situ point spread function (PSF) retrieval is disclosed which includes encoding 3D location of molecules into PSFs, receiving molecule-generated images containing PSFs, segmenting the images into sub-PSFs, initializing template PSFs from a pupil function, determining a maximum normalized cross correlation (NCC) coefficient (NCCmax) between the sub- and template PSFs, associating each of the sub-PSFs with a template PSF based on the NCCmax and storing the sub-PSFs in associated bins, aligning and averaging the binned sub-PSFs, applying a phase retrieval algorithm to the averaged sub-PSFs to update the pupil function, regenerating the template PSFs, repeating until a difference between a new and a prior generation pupil function is below a predetermined threshold, generating in situ PSFs from the last pupil function, and applying a maximum likelihood estimation algorithm based on the in situ PSFs and the sub-PSFs to thereby generate lateral and axial locations of molecules.

Abstract: A method for training an image recognition model includes: obtaining training image sets; obtaining a first predicted probability, a second predicted probability, a third predicted probability, and a fourth predicted probability based on the training image sets by using an initial image recognition model; determining a target loss function according to the first predicted probability, the second predicted probability, the third predicted probability, and the fourth predicted probability; and training the initial image recognition model based on the target loss function, to obtain an image recognition model.

Abstract: An information processing device includes an identification unit that detects a hand region of a person who is in contact with an object included in an input image and identifies an object region that is a region of the object based on a result thereof, a specification unit that specifies an unknown object candidate region that is a candidate for a region of an unknown object from among the object regions identified by the identification unit, a determination unit that determines whether the object is an unknown object based on the unknown object candidate region, and a generation unit that generates, when the object is an unknown object, feedback information for changing an attitude of the unknown object with respect to the person.

Abstract: The invention relates to a method for enhancing images for metrological applications. The method comprises the steps: 1) providing a geometric correction image having an image geometric correctness higher than the processed image geometric correctness and showing at least a part of the scene of interest, and 2) at least partially reducing the loss of initial metrological information in the distorted metrological information by fusing the enhanced image with the geometric correction image.

Abstract: Aspects of the present disclosure include a method for sorting cells of a sample based on an image of a cell in a flow stream. Methods according to certain embodiments include detecting light from a sample having cells in a flow stream, generating an image mask of a cell from the sample and sorting the cell based on the generated image mask. Systems having a processor with memory operably coupled to the processor having instructions stored thereon, which when executed by the processor, cause the processor to generate an image mask of a cell in a sample in a flow stream and to sort the cell based on the generated image mask are also described. Integrated circuit devices (e.g., field programmable gate arrays) having programming for generating an image mask and for determining one or more features of the cell are also provided.

Abstract: Provided in the present application are an image noise reduction method and device, an imaging system, and a non-transitory computer-readable storage medium. The image noise reduction method includes: processing, based on a first deep learning network, an original scanned object image to acquire a noise image corresponding to the original scanned object image; and acquiring a denoised image based on the original scanned object image and the noise image; wherein the first deep learning network is obtained by training based on low signal-to-noise ratio images and high signal-to-noise ratio images.

Abstract: A system and method are provided for collection and testing of a biologic sample. The system and method comprise collecting by a user of a testing device a biologic sample for use with the testing device, assigning correlative values as test results, and receiving the test results at a server disposed on a network. Some aspects further include presenting advertisements and other messages to users through a mobile application operating on a mobile device. These aspects take into account the results of the self-diagnostic test and present different advertisements to the user based on the results of the test.

Abstract: The present disclosure provides methods, apparatus, and storage medium for generating a training sample of a semantic segmentation model. The method includes obtaining an object identifier of a to-be-trained object and a model file corresponding to the object identifier, the model file comprising a three-dimensional model of the to-be-trained object and a skin map set, the skin map set comprising a plurality of solid-color skin maps and a plurality of colorful skin maps; determining transformation models corresponding to the three-dimensional model; generating a plurality of sample image groups corresponding to each transformation model according to the plurality of colorful skin maps and each transformation model; generating a labeled image of the each sample image group according to the object identifier, the plurality of solid-color skin maps, and each transformation model; and generating a training sample of the to-be-trained object according to the labeled image and the each sample image group.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for object detection. In one aspect, a method comprises: obtaining: (i) an image, and (ii) a set of one or more query embeddings, wherein each query embedding represents a respective category of object; processing the image and the set of query embeddings using an object detection neural network to generate object detection data for the image, comprising: processing the image using an image encoding subnetwork of the object detection neural network to generate a set of object embeddings; processing each object embedding using a localization subnetwork to generate localization data defining a corresponding region of the image; and processing: (i) the set of object embeddings, and (ii) the set of query embeddings, using a classification subnetwork to generate, for each object embedding, a respective classification score distribution over the set of query embeddings.

Abstract: A method includes receiving, by a processing unit coupled to a first memory device comprising a first type of media and a second memory device comprising a second type of media, data captured from an imaging device coupled to the processing unit and determining, by the processing unit, characteristics of a workload corresponding to processing of the data. The method further includes writing, by the processing unit, a portion of data associated with the workload to the other of the first memory device or the second memory device based on the determined characteristics of the workload and causing the workload to be executed while at least the portion of the data associated with the workload is written to the other of the first memory device or the second memory device.

Abstract: A system and method to partition pores and throats in 2D or 3D scanned tomographic representations. To achieve partitioning, the methods employ a series of techniques that include the Maximum Inscribed Ball technique, Connected Component Labelling technique, and the 3D Binary Morphology technique.

Abstract: There is provided a method for generating a 3D physical model of a patient specific anatomic feature from 2D medical images. The 2D medical images are uploaded by an end-user via a Web Application and sent to a server. The server processes the 2D medical images and automatically generates a 3D printable model of a patient specific anatomic feature from the 2D medical images using a segmentation technique. The 3D printable model is 3D printed as a 3D physical model such that it represents a 1:1 scale of the patient specific anatomic feature. The method includes the step of automatically identifying the patient specific anatomic feature.

Abstract: A non-transitory computer-readable storage medium storing a set of program instructions for an image processing apparatus is disclosed. The set of program instructions, when executed by a controller of the image processing apparatus, causes the image processing apparatus to: acquire input image data indicating an input image; and input the input image data to a trained machine learning model configured to perform calculation processing on the input image data and to generate output image data corresponding to the input image data. The machine learning model is trained to generate the output image data indicating a first area and a second area. The first area is an area in the input image in which it is preferable to perform first image processing. The second area is an area in the input image in which it is preferable to perform second image processing different from the first image processing.