Abstract: A distribution method includes: distributing, when multiple feature maps exist in an image processing model, to each feature map a neuron through which the feature map passes; filtrating, according to importance of neurons of multiple convolution layers in an image processing model, the neurons to obtain a first result; collecting, according to a position attribute of the each neuron in the first result, statistics on a scale of the feature map corresponding to each neuron to obtain a distribution relationship; the distribution relationship indicates a correspondence between the each feature map and the neuron through which the feature map passes; and distributing, according to the distribution relationship, to the each feature map the neuron through which the feature map passes.

Abstract: Systems and methods are disclosed for grouping cells in a slide image that share a similar target, comprising receiving a digital pathology image corresponding to a tissue specimen, applying a trained machine learning system to the digital pathology image, the trained machine learning system being trained to predict at least one target difference across the tissue specimen, and determining, using the trained machine learning system, one or more predicted clusters, each of the predicted clusters corresponding to a subportion of the tissue specimen associated with a target.

Abstract: A reconstruction method of a three-dimensional (3D) human body model includes: acquiring, by a fully convolutional network (FCN) module, a global UVI map and a local UVI map of a body part according to a human body image (S1); estimating, by a first neural network, a camera parameter and a shape parameter of the human body model based on the global UVI map (S2); extracting, by a second neural network, rotation features of joints of a human body based on the local UVI map (S3); refining, by using a position-aided feature refinement strategy, the rotation features of the joints of the human body to acquire refined rotation features (S4); and estimating, by a third neural network, a pose parameter of the human body model based on the refined rotation features (S5). The reconstruction method achieves accurate and efficient reconstruction of the human body model, and improves robustness of pose estimation.

Abstract: A system for reducing a search area for identifying correspondences identifies an overlap region within a first match frame captured by a match camera. The overlap region includes one or more points of the first match frame that are associated with one or more same portions of an environment as one or more corresponding points of a first reference frame captured by a reference camera. The system obtains a second reference frame captured by the reference camera and a second match frame captured by the match camera. The system identifies a reference camera transformation matrix, and/or a match camera transformation matrix. The system defines a search area within the second match frame based on the overlap region and the reference camera transformation matrix and/or the match camera transformation matrix.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to identify patterns in first high anticipation scenarios based on user identification, wherein high anticipation scenarios include video sequences wherein a first vehicle will be within a specified distance of a first object in a first environment around the first vehicle, wherein user identification is determined by viewing portions of a respective video sequence. A first model including a first deep neural network can be trained to determine second high anticipation scenarios based on the patterns identified in the first high anticipation scenarios and a second model including a second deep neural network can be trained to modify locations and velocities of second objects in the first high anticipation scenarios and output modified high anticipation scenarios.

Abstract: The present disclosure relates to image signal processing methods, apparatus, and devices. One example image signal processing method includes obtaining an image signal, where the image signal is derived based on a sensor signal collected by an image sensor, recognizing, by using a neural network, a scene to which the image signal belongs, determining, by using attribute information of the image signal, whether the scene is accurate, and in response to determining that the scene is accurate, performing enhancement processing on the image signal based on the scene to generate an enhanced image signal.

Abstract: A computer system for verifying hail damage and/or detecting hail fraud includes a processor and a non-transitory, tangible, computer-readable storage medium having instructions stored thereon that, in response to execution by the processor, cause the processor to perform operations including: (i) receiving at least one image of at least a portion of a rooftop; (ii) analyzing the at least one image to identify a plurality of damaged locations; (iii) analyzing damaged locations to determine a distance between each of the damaged locations; and (iv) determining, based upon the analyzing, whether the damaged locations are a result of hail damage by determining the distance between at least some of damaged locations.

Abstract: In an approach for guiding a visually impaired user to position a mobile device appropriately in relation to a screen so that dynamic information on the screen can be reliably extracted and conveyed to the visually impaired user, a processor receives an image captured by a camera of a mobile device. A processor performs object recognition on the image to identify a digital screen and a location of the digital screen in the image. A processor retrieves a template of the digital screen. A processor performs angle-sensitive optical character recognition (OCR) on the location of the digital screen in the image. Responsive to a processor determining text on the digital screen can be extracted, a processor conveys the text to a user. Responsive to a processor determining text on the digital screen cannot be extracted, a processor guides the user to re-orient the mobile device to capture a better image.

Abstract: A ground mark determining method and apparatus are provided. The method includes: obtaining, by the computer device, a point cloud grayscale image, the point cloud grayscale image comprising a road segment map; obtaining, by the computer device, ground mark information from the road segment map and running a mark-extraction network model to extract ground marks in the road segment map, the ground mark information comprising information about the ground marks extracted by the mark-extraction network model, and the ground marks indicating driving information marked on a road segment surface; and determining, by the computer device, a target ground mark from the ground marks according to the ground mark information.

Abstract: Systems and methods determining velocity of an object associated with a three-dimensional (3D) scene may include: a LIDAR system generating two sets of 3D point cloud data of the scene from two consecutive point cloud sweeps; a pillar feature network encoding data of the point cloud data to extract two-dimensional (2D) bird's-eye-view embeddings for each of the point cloud data sets in the form of pseudo images, wherein the 2D bird's-eye-view embeddings for a first of the two point cloud data sets comprises pillar features for the first point cloud data set and the 2D bird's-eye-view embeddings for a second of the two point cloud data sets comprises pillar features for the second point cloud data set; and a feature pyramid network encoding the pillar features and performing a 2D optical flow estimation to estimate the velocity of the object.

Abstract: A target detection method based on the fusion of prior positioning of a millimeter-wave radar and a visual feature includes: simultaneously obtaining, based on the millimeter-wave radar and a vehicle-mounted camera after being calibrated, point cloud data of the millimeter-wave radar and a camera image; performing spatial 3D coordinate transformation on the point cloud data to project transformed point cloud data onto a camera plane; generating a plurality of anchor samples based on projected point cloud data according to a preset anchor strategy, and obtaining a final anchor sample based on a velocity-distance weight of each candidate region; fusing RGB information of the camera image and intensity information of an RCS in the point cloud data to obtain a feature of the final sample; and inputting the feature of the final anchor sample into a detection network to generate category and position information of a target in a scenario.

Abstract: The present disclosure relates to method and system for assessing quality of commodities. An image of bulk commodity is captured and segmented into one or more segmented images for classification into one of predefined categories of commodities. The method and system classify the commodities based on generalized features created from training images. One or more features in the training images are determined and grouped to obtain the generalized features. A feature score and corresponding weightage score of the generalized feature is then determined to estimate a predetermined regression score. Based on the generalized features and predetermined regression score, a likelihood score of the segmented image is determined that enables the classification of the input image to one of the predefined categories of commodities.

Abstract: Systems and methods for generating depth models and depth maps from images obtained from an imaging system are presented. A self-supervised neural network may be capable of regularizing depth information from surface normals. Rather than rely on separate depth and surface normal networks, surface normal information is extracted from the depth information and a smoothness function is applied to the surface normals instead of a depth gradient. Smoothing the surface normal may provide improved representation of environmental structures by both smoothing texture-less areas while preserving sharp boundaries between structures.

Abstract: Computer-implemented methods and systems for image analysis of multiband images of geographic regions are described, including a method by one or more computer executing executable instructions stored in one or more non-transitory, tangible, computer readable media, the method comprising: receiving one or more multiband image of a geographic region, the one or more multiband image having pixels; generating a grey level co-occurrence matrix for the pixels in the one or more multiband image; generating a surface index for the one or more multiband image containing information indicative of a surface type represented by one or more of the pixels in the one or more multiband image; and classifying the pixels of the one or more multiband image into one of a group of predefined land cover classes, based on the surface index in combination with the grey level co-occurrence matrix.

Abstract: An inspection system includes an acquisition unit and a determination unit. The acquisition unit acquires an image representing a surface of an object. The determination unit performs color determination processing. The color determination processing is performed to determine a color of the surface of the object based on a plurality of conditions of reflection. The plurality of conditions of reflection are obtained from the image representing the surface of the object as acquired by the acquisition unit, and have a specular reflection component and a diffuse reflection component at respectively different ratios on the surface of the object.

Abstract: Systems and methods for image classification include receiving imaging data of in-vivo or excised tissue of a patient during a surgical procedure. Local image features are extracted from the imaging data. A vocabulary histogram for the imaging data is computed based on the extracted local image features. A classification of the in-vivo or excised tissue of the patient in the imaging data is determined based on the vocabulary histogram using a trained classifier, which is trained based on a set of sample images with confirmed tissue types.

Abstract: Methods, systems and computer program products for improving ground truth quality for modeling are provided. Aspects include receiving a plurality of data inputs, wherein each of the plurality of data inputs has an associated label. Aspects also include training a model based on the plurality of data inputs. Aspects also include generating a plurality of vector representations corresponding to the plurality of data inputs based on the model. Aspects also include clustering the plurality of vector representations into one or more clusters. Aspects also include identifying at least one anomalous data input based on the one or more clusters. The at least one anomalous data input can be a data input of the plurality of data inputs that is mislabeled, contributes to an ambiguous class structure or is an outlier. Aspects also include outputting a notification that provides an indication of the at least one anomalous data input.

Abstract: Systems and methods are described, and an example method includes a training an artificial intelligence (AI) classifier of electromagnetic (EM) scanned items, including obtaining a training set of sample raw EM scans. The set includes a population of sample in-class raw EM scans, which include blocks of EM sensor data from EM scans of regions having in-class objects, and the set includes a population of sample not-in-class raw EM scans, which include blocks of EM sensor data from EM scan of regions without in-class objects. The example includes applying the AI classifier to sample raw EM scans in the training set, measuring errors in the results, and updating classifier parameters based on the errors, until detecting a training completion state.

Abstract: The disclosure provides a method for training generative adversarial network (GAN), a method for generating images by using GAN, and a computer readable storage medium. The method may train the first generator of the GAN with available training samples belonging to the first type category and share the knowledge learnt by the first generator to the second generator. Accordingly, the second generator may learn to generate (fake) images belonging to the second type category even if there are no available training data during training the second generator.

Abstract: A printed circuit board includes a mounting surface, a transceiver mounted on the mounting surface and configured to process radio-frequency signals, and a radio-frequency module mounted on the mounting surface in communication with the transceiver, the radio-frequency module including an interface including M inputs and N outputs, each of the quantities M and N greater than 1, the interface configured so that each of at least two of the M inputs is coupled to a separate output through a separate switch.