Abstract: Video analytics may be performed on user-triggered video recording to generate metadata that describes what was found in the user-triggered video recording. A subsequent video stream may be analyzed, thereby generating metadata associated with the subsequent video stream. The metadata associated with the subsequent video stream may be compared with the metadata associated with the user-triggered recording. When the metadata associated with the subsequent video stream matches the metadata associated one or more of the user-triggered video recordings in accordance with one or more matching criteria, an inquiry is outputted to an operator soliciting whether a user-triggered video recording should be initiated for the subsequent video stream. The method includes receiving from the operator a response to the inquiry indicating whether to initiate a user-triggered video recording of the subsequent video stream, and if so, initiating a user-triggered video recording of the subsequent video stream.

Abstract: A data processing method includes acquiring video frame data including one or more video frames and audio data of a video, and determining position attribute information of a target object in the acquired one or more video frames, the target object being associated with the audio data. The method also includes acquiring a channel encoding parameter associated with the position attribute information, and performing azimuth enhancement processing on the audio data according to the channel encoding parameter to obtain enhanced audio data. Apparatus and non-transitory computer-readable storage medium counterpart embodiments are also contemplated.

Abstract: Systems and methods for determining a head pose of a user are described. Images captured by a head-mounted camera worn by the user are received, the images comprising captured markers, wherein each captured marker is one of a plurality of markers on a surface that are in the FOV of the head-mounted camera. Stored markers having a predetermined orientation and corresponding to each captured marker are identified from a database, the stored markers comprising coordinates for its corresponding surface marker. An orientation of the captured marker relative to its corresponding stored marker is calculated and the stored marker is obtained. Based on the stored marker and the orientation, an absolute pose of each marker is determined. Based on the absolute pose of each marker, the head pose of the user relative to the surface is calculated.

Abstract: The present invention includes a device for transforming 3D image data such that the transformed image aligns with a 2D fluoroscopic image. The process of finding the transform that aligns the features of the 3D image with the 2D image is called image registration. The fluoroscopic image registration process has at its core a multistage framework that progressively refines a registration result. Each stage leverages the fluoroscopic registration engine which iteratively searches the parameter space defined by the stage. The core registration engine creates a candidate transform, uses it to create a digitally reconstructed radiograph, then compares that to the supplied fluoroscopic image, and the iterative alignment engine either ends or uses it to construct a new candidate transform. Similarity measurement of a candidate transform is based on intensities and any custom data relevant to the registration scenario.

Abstract: The present disclosure provides an image processing method and apparatus, a system, a medium, a device and a driving system, which belong to the field of data processing technologies. The image processing method mainly includes: according to perception target features of each perception target feature class in a plurality of predetermined perception target feature classes, and a pixel requirement of a neural network image perception model corresponding to each perception target feature class in a plurality of predetermined neural network image perception models, performing corresponding adjustment to a current environmental image to obtain a plurality of current environmental feature images; by using the corresponding neural network image perception model, perceiving a perception target in each current environmental feature image to obtain a plurality of perception results; and fusing the plurality of perception results to obtain perception result data of the current environmental image.

Abstract: A method including: obtaining images captured using camera(s), depth maps corresponding to images, and pose information; identifying image segment(s) for first image and second image; determining first relative pose of same object captured in first image; determining second relative pose of same object captured in second image; when same object is in-focus in first image, reprojecting at least image segment(s) of first image; and determining for given camera that captured second image a point spread function as function of optical depth; obtaining third image captured using given camera and third depth map corresponding to third image; and applying extended depth-of-field correction to image segment(s) of third image that is out of focus, by using point spread function.

Abstract: The embodiments of the present disclosure provide a method for radiation therapy guidance. The method may include obtain first surface data of a target object to be treated by using an optical camera, wherein the first surface data may reflect the body surface condition of the target object before a treatment fraction; obtain a historical anatomical image of the target object; and determine whether an anatomical image of the target object needs to be reshot for the treatment fraction based on the historical anatomical image and the first surface data.

Abstract: An image processing apparatus including a tone correction unit configured to generate a plurality of tone correction signals by converting tone characteristics of a plurality of image signals in different wavelength bands, an edge enhancement unit configured to generate an edge enhancement signal by performing predetermined processing with respect to at least one image signal of the plurality of image signals, and a combining unit configured to combine a signal based on the plurality of tone correction signals and the edge enhancement signal is provided. The tone correction unit is configured to generate at least two tone correction signals by converting tone characteristics of at least two image signals of the plurality of image signals while maintaining ratios between pixels respectively corresponding to the at least two image signals.

Abstract: The motion estimation of an anatomical structure may be performed using a machine-learned (ML) model trained based on medical training images of the anatomical structure and corresponding segmentation masks for the anatomical structure. During the training of the ML model, the model may be used to predict a motion field that may indicate a change between a first training image and a second training image, and to transform the first training image and a corresponding first segmentation mask based on the motion field. The parameters of the ML model may then be adjusted to maintain a correspondence between the transformed first training image and the second training image and between the transformed first segmentation mask or a second segmentation mask associated with the second training image. The correspondence may be assessed based on at least a boundary region shared by the anatomical structure and one or more other anatomical structures.

Abstract: The present invention discloses an instrument recognition method based on an improved U2 network. The method includes: replacing common convolution of each layer with grouped convolution Grouped Conv on a basis of an RSU, and segmenting a dial plate and a pointer by using the network; performing noise reduction processing on the scale value array by using a mean filter; and determining a position of a scale value corresponding to the pointer by using a peak value, and outputting a reading according to the scale value and preset data. A pointer-type instrument is segmented by using the improved U2 network, and automatic reading of the obtained dial plate is implemented by using a conventional computer vision method. Therefore, compared with manual reading, the method has advantages of high precision, high reliability, fast reading, low costs, and the like, and working efficiency can be greatly improved.

Abstract: Provided are computing systems, methods, and platforms for using machine-learned models to generate a depth map. The operations can include projecting, using a dot illuminator, near-infrared (NIR) dots on the scene. The NIR dots can have a uniform pattern. Additionally, the operations can include capturing, using a single NIR camera, the projected NIR dots on the scene. Moreover, the operations can include generating a dot image based on the captured NIR dots on the scene. Furthermore, the operations can include processing the dot image with a machine-learned model to generate a depth map of the scene. Subsequently, the operations can further include evaluating the generated depth map of the scene and a ground truth depth map, and performing an action based on the evaluation.

Abstract: Example implementations include a method, apparatus and computer-readable medium for detecting an object in an image, by applying, on a first input image, a first object detector configured to generate a bounding box around the object in the first input image. The implementations further include identifying a first bounding box having a first bounding box size around the object in the first input image and determining that the first bounding box size is less than a threshold size. Additionally, the implementations further include generating a second input image by cropping an area of the original image corresponding to the first bounding box and identifying, by applying a second object detector on the second input image, a second bounding box around the object in the second input image. Additionally, the implementations further include performing an image analysis function on the object using information from the second bounding box.

Abstract: A computer-implemented method, according to one embodiment, includes identifying, in a reference video of a production process of a product, a discrete and non-overlapping set of first tasks. The first tasks define at least a first sub-process. The method further includes analyzing a live video of the production process to identify frames of the live video that include second tasks that define a second sub-process, and analyzing the frames of the live video for determining whether a match exists between the first tasks and the second tasks. In response to a determination that the match does not exist, an alert that a deviation is present in the production process depicted in the live video is output.

Abstract: An image processing method is provided. The image processing method is applied to an image signal processor. The image processing method includes the stages detailed in the following paragraph. Binned-Bayer raw data and binned-white raw data are received from an image sensor with an RGBW color filter array. The binned-Bayer raw data include information of red, green, and blue channels for a plurality of pixels. The binned-white raw data include luminance information for the pixels. The pixels are rendered based on both the binned-Bayer raw data and the binned-white raw data.

Abstract: The present invention discloses a PET image reconstruction method based on Swin-Transformer regularization, the reconstruction model adopted is composed of several iterative modules, each iterative module consists of three parts: a EM iterative layer, a Swin-transformer based regularization layer, a pixel to pixel image fusion layer, the regularization layer is used to learn the prior information representing the image, comprising a convolution kernel used to extract the shallow features of the image, a Swin-Transformer layer used to extract the deep features of the image, and a convolution layer and a residual connection are used to fuse deep and shallow features. The image fusion layer fuses the results of EM iteration and regularization. The invention can reconstruct high quality PET images from Sinogram projection data, and greatly reduces the noise level of PET images while retaining the structural information.

Abstract: Systems and methods include obtaining point cloud data representing a point cloud; selecting a subset of the point cloud data based at least in part on a contour metric; grouping sets of points of the subset of the point cloud into one or more clusters based at least in part on one or more distance metrics; for a cluster that satisfies one or more cluster size criteria based on dimensions of a calibration standard, determining whether a distribution of signal intensities of points of the cluster satisfies a distribution criterion; based on a determination that the distribution of signal intensities of points satisfies the distribution criterion, determining boundaries of a region that represents the calibration standard; and storing data identifying a set of points of the point cloud that correspond to the calibration standard.

Abstract: A method for magnetic resonance image quality assessment is provided. The method includes acquiring a magnetic resonance image; processing the magnetic resonance image to extract image parameters associated with a preset metric; and determining, on the basis of the image parameters, an image quality assessment result associated with the corresponding metric.

Abstract: A method, a device, and a non-transitory computer readable medium for detecting label on curved container, includes obtaining a first image, the first image comprising a label on the curved container; obtaining a plurality of reference points in the first image; obtaining a first width of the curved container and first position data of the label according to the plurality of reference points; determining whether the curved container is rotated according to the first width and a predetermined standard width; when the curved container rotated being determined, predicting a second width of the curved container according to the first position data and a predetermined regression equation; and determining whether the label is offset according to the first width and the second width.

Abstract: Examples provide a video security system including an electronic processor configured to detect an anomaly by obtaining video data captured using an image sensor and determining a number of people detected in each of a plurality of frames of the video data. For each moment t, the electronic processor determines an average number of people n over a time interval (t-?t1, t), where ?t1 is greater than or equal to the moment t. For each non-overlapping time interval ?t1, the electronic processor updates a histogram based on the average number of people for the time interval ?t1. In response to determining, using the histogram, that the probability ? of the average number of people n over the time interval (t-?t1, t) is less than a threshold probability ?*, the electronic processor generates an alert indicating that an anomalous crowd size is detected.

Abstract: A system for enabling a plurality of different artificial intelligence models to be applied to image data acquired from one or more inspection systems. The system is configured to receive an instruction from an operator workstation to apply one of a plurality of different artificial intelligence (AI) models to an image representative of cargo. Further, an application program interface specific to that one AI model, is acquired. Further, the image is submitted to the AI model using the application program interface and data representative of a degree to which the cargo in the image corresponds to an expected type of cargo based upon descriptive data associated with the cargo, is received from the AI model. Thereafter, the data representative of the degree to which the cargo in the image corresponds to the expected type of cargo is transmitted to the operator workstation, to be displayed to an operator.