Abstract: A method of labeling data and training a model is provided. The method includes obtaining a set of images. The set of images includes a first subset and a second subset. The first subset is associated with a first set of labels. The method also includes generating a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model. The method further includes generating second machine learning model. The second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of labels, and the second set of labels. The second set of labels is updated based on one or more inferences generated by the second machine learning model.

Abstract: A vision-based target tracking method, a system, an equipment, and a storage medium. The method includes: acquiring face positions and body positions in a current image frame; acquiring a current position of a time-series tracking target by a time-series tracking algorithm; determining, when a face matching the face of a to-be-tracked target does not exist outside a region corresponding to the current position of the time-series tracking target, whether the time-series tracking target is in an intersecting state; determining whether a body matching the body of the to-be-tracked target in other persons intersecting the times-series tracking target when in the intersecting state, or using the current position of the time-series tracking target as a current target position if not in the intersecting state; and taking the current target position as the position of the to-be-tracked target in the previous frame, and returning to the above steps.

Abstract: A method for determining a salient object in an image based on superpixel analysis consists of initializing superpixels using SLIC algorithm and merging of adjacent superpixels that have similar color distribution. Then, calculate the spatial and color distribution correlation between the superpixels after being merged. Combined with the statistics of the occupancy rate, the distance to the original image center, and the global contrast of the superpixels, calculate the saliency evaluation vector for each superpixel. Finally, interpolate the saliency for each pixel in the superpixel.

Abstract: Embodiments of the present invention relates to a stereo matching method, an image processing chip and a mobile vehicle. The stereo matching method includes: calculating aggregate cost values between all reference pixels and a target pixel in a preset search region, the reference pixel being a pixel in a reference image, and the target pixel being a pixel in a target image; determining a texture property of the reference image in the search region according to a distribution of the aggregate cost values; and using a method for calculating a disparity value corresponding to the texture property, to obtain a disparity value between the reference image and the target image at a position of the target pixel.

Abstract: Embodiments described herein provide systems, methods, and devices for pre-training a multimodal encoder-decoder (MED) model for vision-language tasks. A method may include encoding, by an image encoder of the MED, an image into an image representation; encoding, by a text encoder of the MED, a text into a text representation; generating, by an image-grounded text encoder of the MED, a multimodal representation based on the image representation and the text; generating, by an image-grounded text decoder of the MED, a predicted text based on the image representation and the text; generating, through an image-text matching (ITM) head, a binary classification indicating whether the image and the text are a match; computing a first loss, ITM loss, and third loss based on the image representation, text representation, binary classification, predicted text and text; jointly updating the MED based on the first loss, the second loss and the third loss.

Abstract: A Smart Optical Character Recognition (SOCR) Trainer comprises software developed for automating Quality Control (QC) using unsupervised machine-learning techniques to analyze, classify, and optimize textual data extracted from an image or PDF document. SOCR Trainer serves as a ‘data treatment’ utility service that can be embedded into data processing workflows (e.g., data pipelines, ETL processes, data versioning repositories, etc.). SOCR Trainer performs a series of automated tests on the quality of images and their respective extracted textual data to determine if the extraction is trustworthy. If deficiencies are detected, SOCR Trainer will analyze certain parameters of the document, perform conditional optimizations, re-perform text extraction, and repeat QA testing until the output meets desired specifications. SOCR Trainer will produce audit files recording the provenance and differences between original documents and enhanced optimized document text.

Abstract: A computer-implemented method can include a training phase and a hemorrhage detection phase. The training phase can include: receiving a first plurality of frames from at least one original computed tomography (CT) scan of a target subject, wherein each frame may or may not include a visual indication of a hemorrhage, and further wherein each frame including a visual indication of a hemorrhage has at least one label associated therewith; and using a fully convolutional neural network (FCN) to train a model by determining, for each of the first plurality of frames, whether at least one sub-portion of the frame includes a visual indication of a hemorrhage and classifying the sub-portion of the frame based on the determining.

Abstract: A monitoring camera includes an imaging unit configured to capture an image of an imaging area, and a processor configured to determine a first intensity and a second intensity to be different from each other. The first intensity indicates an intensity of a noise reduction processing executed on an attention portion in the captured image of the imaging area, and the second intensity indicates an intensity of a noise reduction processing executed on a non-attention portion in the captured image. The first intensity is lower than the second intensity. The processor is configured to execute the noise reduction processing on the attention portion based on the determined first intensity, to execute the noise reduction processing on the non-attention portion based on the determined second intensity, and to output an image after the noise reduction processing.

Abstract: The present disclosure relates to a data processing method, and more specifically, to a digital image processing method to enable a rapid noise-suppressed contrast enhancement in an optical linear or nonlinear microscopy imaging application. The disclosed method digitally mimics a hardware-based feedback-driven adaptive or controlled illumination technique by means of digitally resembling selective laser-on and laser-off states so as to selectively optimize the signal strength and hence the visibility of the weak-intensity morphologies while mostly preventing saturation of the brightest structures.

Abstract: A system for inspecting a battery component includes a heating device configured to heat a surface of the battery component to a selected temperature, an optical-visible imaging device configured to take an optical image of the surface, a thermal imaging device configured to take a thermal image of the surface, and a processor configured to acquire the optical image and the thermal image. The processor is configured to correlate the thermal image with the optical image, identify a feature of interest in at least one of the optical image and the thermal image, determine a geometric characteristic and a temperature characteristic associated with the feature of interest, and determine whether the feature of interest is a defect based on the geometric characteristic and the temperature characteristic.

Abstract: The current invention concerns a method, a computer system, and a computer program product for blood vessel image extraction from ultrasound signals, and a method and apparatus for ultrasound imaging A set of ultrasound signals at successive times in a plurality of spatial points is obtained. The ultrasound signals are decomposed as a sum of terms. Each term consists of a time-independent spatial component image factor and a space-independent temporal component signal. The spatial component images are mutually orthonormal. The temporal component signals are mutually orthogonal. A sharpened spatial image for each of multiple spatial component images is determined. A processed image being a bilinear combination of multiple sharpened spatial images is generated.

Abstract: A system for calibrating a plurality of user sensors includes a calibration management system comprising a processor and a node services module and a plurality of calibration nodes disposed at a plurality of geographical locations. The processor is communicably coupled to the plurality of user sensors, the node services module, and the plurality of calibration nodes and operable to: provide calibration node information to a system associated with one of the plurality of user sensors and cause the calibration node to perform at least one calibration action after providing the calibration information, the calibration action may include performing an environmental measurement and/or providing a calibration reference signal to the user sensor. The processor is also operable to provide calibration results information containing results of the calibration action to the system associated with the one of the plurality of user sensors.

Abstract: A system and method for providing a decision basis for controlling a robotic arm to perform at least one action relating to a milk-producing animal, wherein a camera registers three-dimensional image data of a milking location that includes a reference object and a rotating platform upon which an animal stands with its hind legs facing the camera, and a control unit checks if an entry window for a robotic arm can be found in the image data by searching for the reference object in the image data, and if the reference object is found the control unit searches for an acceptable obstacle-free volume in the image data, this volume being located within an allowed space relative to the reference object.

Abstract: The present invention provides an intelligent and a flexible video enhancement system with improved visual quality of the video for better user experience. The video enhancement system enhances various types of videos and provides a user with better visual experience based on video corresponding features and personal preferences. The video enhancement system includes a dynamic analysis and adaptive processing techniques for enhancing video quality for seamless user experience.

Abstract: A method and a system for detecting a scene text may include extracting a first feature map for a scene image input based on a convolutional neural network, and delivering the first feature map to a sequential deformation module; obtaining sampled feature maps corresponding to sampling positions by performing iterative sampling for the first feature map, obtaining a second feature map by performing a concatenation operation in deep learning according to a channel dimension for the first feature map and the sampled feature maps; obtaining a third feature map by performing a feature aggregation operation for the second feature map in the channel dimension, and delivering the third feature map to the object detection baseline network; and performing text area candidate box extraction for the third feature map and obtaining a text area prediction result as a scene text detection result through regression fitting.

Abstract: Provided is a method for locating a tumor. The method includes: performing image registration on a projection image of the tumor and a first standard image to acquire a first offset; generating a second standard image based on the first offset; performing image registration on the projection image and the second standard image to acquire a second offset; and updating the second standard image based on the second offset and executing the operation of image registration on the projection image and the second standard image again in response to the second offset satisfying a virtual re-sampling condition; or outputting an accumulated offset in response to the second offset not satisfying the virtual re-sampling condition, wherein the accumulated offset is a sum of the first offset and the second offset acquired by executing the operation of image registration.

Abstract: A method and computing system for facilitating edge detection are presented. The computing system may include at least one processing circuit configured to receive image information representing a group of objects in a camera field of view, and to identify, from the image information, a plurality of candidate edges associated with the group of objects. The at least one processing circuit may further determine, when the plurality of candidate edges include a first candidate edge which is formed based on a border between a first image region and a second image region, whether the image information satisfies a defined darkness condition at the first candidate edge. The at least one processing circuit may further select a subset of the plurality of candidate edges to form a selected subset of candidate edges for representing the physical edges of the group of objects.

Abstract: Techniques are described for using computing devices to perform automated operations to generate mapping information via analysis of visual data of photos of a defined area, and for using the generated mapping information in further automated manners, including to display the generated mapping information via various types of visualizations corresponding graphical user interfaces. In some situations, the defined area includes an interior of a multi-room building, and the generated information includes at least a partial floor plan and/or other modeled representation of the building—in addition, the generating may be further performed without having measured depth information about distances from the photos' acquisition locations to walls or other objects in the surrounding building. The generated floor plan and/or other mapping-related information may be further used in various manners, including for controlling navigation of devices (e.g., autonomous vehicles), etc.

Abstract: Described herein are systems, methods and instrumentalities associated with automatic assessment of aneurysms. An automatic aneurysm assessment system or apparatus may be configured to obtain, e.g., using a pre-trained artificial neural network, strain values associated one or more locations of a human heart and one or more cardiac phases of the human heart and derive a representation (e.g., a 2D matrix) of the strain values across time and/or space. The system or apparatus may determine, based on the derived representation of the strain values, respective strain patterns associated with the one or more locations of the human heart and further determine whether the one or more locations are aneurysm locations by comparing the automatically determined strain patterns with predetermined normal strain patterns of the heart and determining the presence or risk of aneurysms based on the comparison.

Abstract: In the method, a predetermined first number of sensed positions (Pi) of the object (20) are provided, the positions being provided with respect to a vehicle coordinate system (K) of the vehicle (10). Depending on a predetermined second number of the provided positions, a second coordinate system (K?) is determined. The provided positions are transformed into the second coordinate system (K?) and a function is determined which approximates a course of the provided positions in the second coordinate system (K?). For a predetermined third number of sampling points of the function, at least one analytical property of the function is determined at the respective sampling point. The coordinates and the at least one analytical property of the respective sampling point are transformed into the vehicle coordinate system (K).