Abstract: Upon a triggering event, frames of a video feed are analyzed for activity or motion. One frame or a subset of the frames are selected as a keyframe/keyframes based on one or measures associated with one or more metrics derived from the frames. The keyframe/keyframes optimally reflects an activity or degree of motion across the frames. The keyframe/keyframes are then provided as input to a computer vision or machine learning application to improve the efficiency, accuracy, and/or precision of results produced by or decisions made by the application.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for adapting generative neural networks to target domains utilizing an image translation neural network. In particular, in one or more embodiments, the disclosed systems utilize an image translation neural network to translate target results to a source domain for input in target neural network adaptation. For instance, in some embodiments, the disclosed systems compare a translated target result with a source result from a pretrained source generative neural network to adjust parameters of a target generative neural network to produce results corresponding in features to source results and corresponding in style to the target domain.

Abstract: An image processing method for a video processor, for generating an extrapolated frame according to a previous frame and a current frame, includes steps of: projecting a plurality of motion vectors (MVs) to the extrapolated frame subsequent to the current frame; determining whether a block of the extrapolated frame is projected by at least two of the MVs; selecting at least two candidate MVs from the MVs projected to the block when the block is projected by at least two of the MVs; calculating a blended MV which is a mixture of the at least two candidate MVs, and projecting the blended MV to the previous frame; obtaining a reference MV corresponding to position of the previous frame projected by the blended MV; and comparing the reference MV with the at least two candidate MVs, to select a final MV for the block from the at least two candidate MVs.

Abstract: An evaluation method and system for the corrosion degree of an absorbable stent.

Abstract: Disclosed are a face search method and device, an apparatus, and a computer-readable storage medium. According to the data volume of feature information in image feature databases currently to be searched, a user may flexibly select a technique of matching search feature information against feature information in a corresponding feature database, and/or select a technique of first performing initial matching for the search feature information against indices corresponding to feature information in a corresponding feature database, and then performing exact matching for the search feature information against feature information corresponding to a successfully matched index obtained from the initial matching.

Abstract: Techniques are disclosed for a covariance model that may generate observation covariances based on observation data of object detections. Techniques may include determining observation data for an object detection of an object represented in sensor data, determining that track data of a track is associated with the object, and inputting the observation data associated with the object detection into a machine-learned model configured to output a covariance (a covariance model). The covariance model may output one or more observation covariance values for the observation data. In some examples, the techniques may include determining updated track data based on the track data, the one or more observation covariance values, and the observation data.

Abstract: An information processing apparatus calculates an accurate index for the growth state of a plant without increasing a user load. The apparatus includes a receiver that receives an input of a captured image obtained by capturing an image of a plant cultivation area including a rectangular area in a longitudinal direction of the rectangular area, a conversion factor calculator that calculates conversion factors at positions in a depth direction in the captured image by determining a pixel count in a width direction of the rectangular area in the captured image input into the receiver at each position in the depth direction and dividing an actual width of the rectangular area by the determined pixel count, and an index calculator that calculates an index for a growth state of the plant by analyzing the captured image using the conversion factors at the positions calculated by the conversion factor calculator.

Abstract: An image processing apparatus and a method of operating the same are provided. The method includes: a memory storing one or more instructions; and a processor configured to execute the one or more instructions stored in the memory to obtain first frequency coefficient information by converting a first image into a frequency domain in units of blocks having a preset size, obtain correlation information indicating a correlation between at least one block of the first frequency coefficient information and a first kernel, generate a weight corresponding to the first frequency coefficient information based on the correlation information, generate second frequency coefficient information by rearranging coefficients included in the first frequency coefficient information, wherein the one or more of the coefficients having a same frequency is arranged into a same group, and obtain quality information of the first image based on the weight and the second frequency coefficient information.

Abstract: Disclosed herein are system, method, and computer program product embodiments for verifying a document. An embodiment operates by receiving an initial image of a document and identifying a first imperfection on a feature of the document. Thereafter, the feature of the initial image is segmented into a first and second partially overlapping segment corresponding to a first and second portion of the feature, respectively. Subsequently, the feature's first portion is determined to be free from the first imperfection, and the first imperfection is identified in the feature's second segment. After receiving a second image of the document including the feature's second portion, the feature's second portion in the second image is determined to be free from a second imperfection. As a result, an image of the feature is created based on the first and second portions of the feature derived from the initial and second images, respectively.

Abstract: A system and method allows a light colored image of an object such as a check to be detected and captured on a light colored background for uploading to a server for processing. Detection involves detecting edges of objects on the image, drawing a rectangle around the detected edges, testing for an aspect ratio of the rectangle within an approved range, testing for the rectangle being outside of a margin of the image and being a certain percentage of the image, and testing for blur within a tolerable range.

Abstract: A method of encoding point cloud data using a planar coding mode is disclosed. The planar coding mode may be signaled using a planar mode flag to signal that a current volume is planar. A volume is planar if all of its occupied child nodes are on one side of a plane bisecting the volume. A planar position flag may signal which side of the volume is occupied. Volume data for already-coded occupied volumes of the point cloud is tracked using a data structure stored in memory. Entropy coding may be used to code the planar mode flag and/or the planar position flag. Context determination for coding may take into account a distance between the volume and a closest already-coded occupied volume among those tracked already-coded occupied volumes that have a same index in the data structure as the current volume.

Abstract: A computer-implemented method of recognition of actions performed by individuals includes: by one or more processors, obtaining images including at least a portion of an individual; by the one or more processors, based on the images, generating implicit representations of poses of the individual in the images; and by the one or more processors, determining an action performed by the individual and captured in the images by classifying the implicit representations of the poses of the individual.

Abstract: A determination system includes an imaging unit that captures an image of a gasket already used for flange fastening and an analyzing unit that analyzes, based on the image captured, a contact mark formed on the gasket as a result of contact between the gasket and a pair of flanges when the flanges are fastened through the gasket. The analyzing unit includes a detection unit that detects, in the captured image, a contact mark region indicative of the contact mark and enclosed with a first line and a second line, a distance calculation unit that calculates a distance between the first line and the second line, and a state determination unit that determines, based on at least the distance, whether the flanges are properly fastened through the gasket.

Abstract: An information processing apparatus calculates an accurate index for the growth state of a plant without increasing a user load. The apparatus includes a receiver that receives an input of a captured image obtained by capturing an image of a plant cultivation area including a rectangular area in a longitudinal direction of the rectangular area, a conversion factor calculator that calculates conversion factors at positions in a depth direction in the captured image by determining a pixel count in a width direction of the rectangular area in the captured image input into the receiver at each position in the depth direction and dividing an actual width of the rectangular area by the determined pixel count, and an index calculator that calculates an index for a growth state of the plant by analyzing the captured image using the conversion factors at the positions calculated by the conversion factor calculator.

Abstract: A method for generating a video is described. The method includes: generating an initial output video including multiple frames, each of the frames having multiple channels; identifying a partitioning of the initial output video into a set of channel slices that are indexed according to a particular slice order, each channel slice being a down sampling of a channel stack from a set of channel stacks; initializing, for each channel stack in the set of channel stacks, a set of fully-generated channel slices; repeatedly processing, using an encoder and a decoder, a current output video to generate a next fully-generated channel slice to be added to the current set of fully-generated channel slices; generating, for each channel index, a respective fully-generated channel stack using the respective fully generated channel slices; and generating a fully-generated output video using the fully-generated channel stacks.

Abstract: Mechanisms are provided to implement a neural flow attestation engine and perform computer model execution integrity verification based on neural flows. Input data is input to a trained computer model that includes a plurality of layers of neurons. The neural flow attestation engine records, for a set of input data instances in the input data, an output class generated by the trained computer model and a neural flow through the plurality of layers of neurons to thereby generate recorded neural flows. The trained computer model is deployed to a computing platform, and the neural flow attestation engine verifies the execution integrity of the deployed trained computer model based on a runtime neural flow of the deployed trained computer model and the recorded neural flows.

Abstract: Systems and methods for adapting a neural network model on a hardware platform. An example method includes obtaining neural network model information comprising decision points associated with a neural network, with one or more first decision points being associated with a layout of the neural network. Platform information associated with a hardware platform for which the neural network model information is to be adapted is accessed. Constraints associated with adapting the neural network model information to the hardware platform are determined based on the platform information, with a first constraint being associated with a processing resource of the hardware platform and with a second constraint being associated with a performance metric. A candidate configuration for the neural network is generated via execution of a satisfiability solver based on the constraints, with the candidate configuration assigns values to the plurality of decision points.

Abstract: A method of masking in an output image stream includes receiving an input image stream capturing a scene, processing the input image stream to generate the output image stream, including using a detector to detect objects in the scene and a tracker to track objects in the scene based on information provided by the detector, and further including to generate a particular output image of the output image stream by checking whether there exists a particular area in the scene in which an evaluation of a historical performance of the detector and/or tracker fulfills at least one condition, and to, if confirming that such a particular area exists, mask the particular area of the scene in the particular output image. A corresponding device, computer program, and computer program product are also provided.

Abstract: A method for situation awareness is provided. The method comprises: preparing a neural network trained by a learning set, wherein the learning set includes a plurality of maritime images and maritime information including object type information which includes a first type index for a vessel, a second type index for a water surface and a third type index for a ground surface, and distance level information which includes a first level index indicating that a distance is undefined, a second level index indicating a first distance range and a third level index indicating a second distance range greater than the first distance range; obtaining a target maritime image generated from a camera; and determining a distance of a target vessel based on the distance level index of the maritime information being outputted from the neural network which receives the target maritime image and having the first type index.

Abstract: An image recognition method includes: obtaining a target three-dimensional (3D) image; inputting the target 3D image to a first recognition model; and obtaining the image type of the target 3D image outputted by the first recognition model. The first recognition model is configured to perform image recognition on the target 3D image to obtain an image type of the target 3D image. A convolutional block of the first recognition model is the same as a convolutional block of a second recognition model, and configured to perform image recognition on the target 3D image. The second recognition model is obtained by training an original recognition model using a target training sample, the target training sample including cubes obtained by rotating and sorting N target cubes obtained from a 3D sample image, N being a natural number greater than 1.