Abstract: Magnetic resonance imaging (MRI) image reconstruction using machine learning is described. A variational or unrolled deep neural network can be used in the context of an iterative optimization. In particular, a regularization operation can be based on a deep neural network. The deep neural network can take, as an input, an aliasing data structure being indicative of aliasing artifacts in one or prior images of the iterative optimization. The deep neural networks can be trained to suppress aliasing artifacts.

Abstract: A method comprising: receiving a dataset comprising (i) a plurality of images, and (ii) a set of classes associated with one or more objects in the images; selecting at least one image from the dataset; applying one or more transformations to the selected image, to create a set of transformed images, wherein each of the transformed images includes a representation of at least one of the objects; annotating at least some of the objects in the selected image and at least some of the transformed images, wherein the annotating comprises assigning each of the one or more objects to one of the classes; and calculating an ambiguity score with respect to at least one pair of classes in the set of classes, based, at least in part, on a number of times the annotating assigned a one of the objects to both of the classes in the pair.

Abstract: A sequence of images generated at respective times by one or more sensors configured to sense an environment through which objects are moving relative to the one or more sensors is received. A message passing graph having a multiplicity of layers associated with the sequence of images is constructed. A neural network supported by the message passing graph is trained. The training includes performing a pass through the message passing graph in a forward direction including by adding a new feature node based on a feature detection and a new edge node and performing a pass through the message passing graph in a backward direction, including by updating at least one edge node of the message passing graph. Multiple features are tracked through the sequence of images, including passing messages through the message passing graph.

Abstract: A method for measuring photosynthetically active radiation (PAR), and particularly fraction-absorbed PAR (faPAR) measurements, which shows a relationship between how much energy is available compared with how much energy is actually used, or absorbed, by plants in a region of interest, can include calibrating multiband image data to generate reflectance-calibrated values for each band of the multiband image data. The weighted reflectance data of the bands can be combined, along with down-welling light data captured at the time of the multi-spectral images to generate a faPAR image. In some cases, faPAR is generated using a ratio of up-welling PAR (uPAR) to down-welling PAR (dPAR). The dPAR can be generated using partially-reflectance-calibrated data and fully-reflectance-calibrated data and the uPAR can be generated using the partially-reflectance-calibrated data.

Abstract: Disclosed is a tracking and identification method for multiple vessel targets, devices, electronic device, and storage media. The method comprises: determining the current position of the vessel based on effective AIS data, and projecting it into an image to obtain the visual motion trajectory of the vessel; obtaining target detection boxes corresponding to multiple vessels based on video surveillance data; determining an occluded area based on the target detection boxes of multiple vessels at the previous time, determining the predicted detection box of the occluded area, and loading the appearance features of the predicted detection boxes as the appearance features extracted at the last time before occlusion; integrating effective AIS data from multiple vessels into video surveillance data to determine the vessel's identity. This disclosure can solve the problem of anti-occlusion tracking for vessels in complex vessel navigation scenarios such as severe occlusion and complete occlusion.

Abstract: A method for automated determination of a growth rate of an object in 3D data sets is described wherein the method may comprise: a first trained 3D detection deep neural network (DNN) determining one or more first VOIs in a current 3D data set and second VOIs in prior 3D data set, a VOI being associated with an abnormality; a registration algorithm, preferably a registration algorithm based on a trained 3D registration DNN, determining a mapping between the one or more first and second VOIs, the mapping providing for a first VOI in the current 3D data set a corresponding second VOI in the prior 3D data set; a second trained 3D segmentation DNN segmenting voxels of a first VOI into first voxels representing the abnormality and voxels of a corresponding second VOI into second voxels representing the abnormality; and, determining a first volume of the abnormality on the basis of the first voxels and a second volume of the abnormality on the basis of the second voxels and using the first and second volume to dete

Abstract: A system and method for training a neural network to denoise images. One noise realization is paired to an ensemble of training-ready noise realizations, and fed into a neural network for training. Training datasets can also be retrospectively generated based on existing patient studies to increase the number of training datasets.

Abstract: A method of tracking an object across a stream of images comprises determining a region of interest (ROI) bounding the object in an initial frame of an image stream. A HOG map is provided for the ROI by: dividing the ROI into an array of M×N cells, each cell comprising a plurality of image pixels; and determining a HOG for each of the cells. The HOG map is stored as indicative of the features of the object. Subsequent frames are acquired from the stream of images. The frames are scanned ROI by ROI to identify a candidate ROI having a HOG map best matching the stored HOG map features. If the match meets a threshold, the stored HOG map indicative of the features of the object is updated according to the HOG map for the best matching candidate ROI.

Abstract: An image restoration method and apparatus are provided. The image restoration method includes acquiring a target image, and acquiring a restoration image of the target image from an image restoration model to which the target image and pixel position information of the target image are input.

Abstract: An evaluation system is a system configured to evaluate coverage of an evaluation target by using a captured image of the evaluation target. The evaluation system includes: an image acquisition unit configured to acquire the captured image; a correction unit configured to generate an evaluation image by correcting the captured image; an evaluation unit configured to evaluate the coverage based on the evaluation image; and an output unit configured to output a result of the evaluation carried out by the evaluation unit, wherein the correction unit extracts an evaluation region from the captured image based on the size of a dent region included in the captured image and generates the evaluation image based on the evaluation region, and the dent region is an image of a dent formed on the evaluation target.

Abstract: This application discloses a method for detecting a target in a video, performed by a computing device. The method includes applying a target detection operation to a first frame in the video, to determine a first target detection result of the target in the first frame; applying a target tracking operation to a second frame after the first frame in the video, to determine changes of the target between the first frame and the second frame; and determining a second target detection result of the target in the second frame according to the first target detection result and the changes of the target between the first frame and the second frame.

Abstract: Various methods and systems are described for obtaining at least one CTA perfusion functional map from Time Resolved Helical CTA (TRH-CTA) image data. At least one processor may be configured to preprocess the TRH-CTA helical image data to generate preprocessed TRH-CTA helical image data; generate time density curve data for a plurality of voxels from the preprocessed TRH-CTA helical image data for an axial imaging slice, where the time density curve data comprise intensity values for different phases of the preprocessed TRH-CTA helical image data arranged sequentially in time; generate at least one perfusion functional map for the axial imaging slice by at least one of: (1) applying at least one mapping function to different phases of the time density curve data corresponding to the axial imaging slice; (2) applying a deconvolution method to the time density curve data; and (3) applying a non-deconvolution method to the time density curve data; and perform spatial filtering on the perfusion functional map.

Abstract: A method for motion recognition and embedding is disclosed. The method may include receiving a plurality of frames of an input video for extracting a feature vector of a motion in the plurality of frames, generating a plurality of sets of one or more motion component bits based on the feature vector and a plurality of classifiers, the plurality of sets corresponding to the plurality of classifiers, each set of one or more motion component bits representing a physical or mechanical attribute of the motion; and generating a motion code for a machine to execute the motion by combining the plurality of sets of one or more motion component bits. Other aspects, embodiments, and features are also claimed and described.

Abstract: A method for recognizing at least one object in at least one input image. In the method, a template image of the object is processed by a first convolutional neural network (CNN) to form at least one template feature map; the input image is processed by a second CNN to form at least one input feature map; the at least one template feature map is compared to the at least one input feature map; it is evaluated from the result of the comparison whether and possibly at which position the object is contained in the input image, the convolutional neural networks each containing multiple convolutional layers, and at least one of the convolutional layers being at least partially formed from at least two filters, which are convertible into one another by a scaling operation.

Abstract: Disclosed are methods and systems that include obtaining at least one image of a dendritic structure, analyzing the at least one image to identify one or more features associated with the dendritic structure, and determining a numerical value associated with the dendritic structure based on the one or more features.

Abstract: An asset identification and tracking system includes one or more monitoring units configured to monitor at least one designated area. Each of the monitoring units includes an imaging device and one or more processors. The imaging device is configured to generate image data depicting one or more mobile assets that move through the at least one designated area. The one or more processors are operably coupled to the imaging device and configured to analyze the image data to detect and decipher one or more identifiers that are displayed on a particular mobile asset of the one or more mobile assets that move through the at least one designated area. The one or more processors are further configured to generate a detection message that includes the one or more identifiers for communication to an asset control system.

Abstract: An electronic device and a method for controlling the electronic device are provided. The method for controlling the electronic device includes, based on an occurrence of an event for outputting information being determined, obtaining data for determining a context corresponding to the electronic device, inputting the obtained data to a first model trained by an artificial intelligence algorithm and obtaining information about a person located in a vicinity of the electronic device, inputting the obtained information about the person and information about the event to a second model trained by an artificial intelligence algorithm and obtaining output information corresponding to the event, and providing the obtained output information.

Abstract: Systems and methods are provided for detecting one or more anomalous events in video. Histogram-based noise cleansing, higher-order deep convolutional neural network-based feature extraction, instance segmentation, instance summation, difference calculation, and normalization can be used. Human-in-loop systems and methods can facilitate human decisions for anomaly detection. The decision of an anomaly event can be made by, for example, the instance difference value(s).

Abstract: A method for generating a monitoring image. The method includes: providing an image sequence of the surroundings to be monitored with the aid of an imaging system; determining at least one monitoring area and at least one periphery area of at least one image of the image sequence with the aid of a learning-based semantic segmentation method; compressing the monitoring area of the at least one image of the image sequence with a first compression quality; and compressing the periphery area of the at least one image of the image sequence with a second compression quality to generate the compressed monitoring image, the second compression quality being lower than the first compression quality.

Abstract: Data may be abstracted and/or masked prior to being provided to a machine learning model for training. A machine learning model may provide a confidence level associated with a result. If the confidence level is too high, the machine learning model or an application including the machine learning model may refrain from providing the result as an output. In some examples, the machine learning model may provide a “second best” result that has an acceptable confidence level. In other examples, an error signal may be provided as the output. In accordance with examples of the present disclosure, data may be abstracted and/or masked prior to being provided to a machine learning model for training and confidence levels of results of the trained machine learning model may be used to determine when a result should be withheld.