Abstract: A method, data processing apparatus, and computer code for identifying road intersections includes providing location data obtained from at least one vehicle's trajectory, wherein the location data may include geographical data within a geographical perimeter. The method includes determining node vectors by applying a geographical descriptor model on a target location included in the geographical perimeter. The geographical descriptor model includes a plurality of multiscale node descriptors including a target multiscale descriptor and neighboring multiscale descriptors. Each of the plurality of multiscale node descriptors includes at least two shape descriptors of different geographical resolution. Each of the neighboring locations is at a respective geographical distance from the target location. The node vectors may be respectively determined for each of the plurality of multiscale node descriptors.

Abstract: An example apparatus for selecting priors includes a training set receiver to receive a training dataset. The apparatus includes a prior generator to generate a set of redundant priors based on the training dataset. The apparatus includes an intermediate trainer to train a detection CNN using the set of redundant priors. The apparatus includes a score and location receiver to send all training samples of the training dataset to the trained detection CNN and receive responses for all of the redundant priors in the set of redundant priors. The apparatus includes a subset selector to select a subset of the set of redundant priors based on the responses.

Abstract: A system for flexible regularization and adaptable scaling of an artificial neural network is provided.

Abstract: A computer-implemented method, system and computer program product for embedding contextual information in an image or video frames. A generative adversarial network (GAN) is trained to provide contextual information to be embedded in an image or video frames, where the contextual information includes text, sound and/or video frames that provides context to the image or video frames. After training the GAN, an image or video frames are received to be embedded with contextual information if necessary. Features are then extracted from the received image/video frames. An image(s) or video frame(s) are identified in a database using the GAN associated with features with a similarity to the extracted features of the received image/video frames that exceeds a threshold value. Such identified images and/or video frames are associated with “references” containing contextual information which are extracted. The received image/video frames are then augmented with the extracted references to provide context.

Abstract: Disclosed is a method for recognizing distribution network equipment based on Raspberry Pi multi-scale feature fusion. The method includes obtaining an initial sample data set; constructing an object detection network composed of EfficientNe-B0 backbone network, multi-scale feature fusion module and a regression classification prediction head; training the object detection network by taking the initial sample data set as a training sample; finally, detecting inspection pictures by using a the trained object detection network. A light-weight EfficientNet-B0 backbone network feature extraction method obtains more features of objects. Meanwhile, an introduction of multi-scale feature fusion better adapts to small object detection, and a light-weight y_pred regression classification detection head is effectively deployed and realized in Raspberry Pi embedded equipment with tight resources and limited computing power.

Abstract: A method for training a model, the method including: defining a primary model for identifying a class of input data based on a first characteristic of the input data; defining a secondary model for detecting a change to a second characteristic between multiple input data captured at different times; defining a forward link from an output of an intermediate layer of the secondary model to an input of an intermediate layer of the primary model; and training the primary model and the secondary model in parallel based on a training set of input data.

Abstract: Various methods are provided for training and subsequently utilizing a convolutional neural network (CNN) to detect small pedestrians (e.g., pedestrians located away a large distance). One example method may comprise performing a first training stage in which a first CNN is trained to detect objects of a first size, the first CNN trained using a first set of images comprised of objects of the first size, and configured to output a first set of parameters, performing a second training stage in which a second CNN is trained using a second set of images, the second set of images comprising objects of a second size, and the first CNN is initialized with the first set of parameters and is re-trained using the second set of images, and determining parameters of the first CNN by minimizing error between the first CNN and the second CNN.

Abstract: A method and apparatus for performing preemptive rendering of a 3D dataset is disclosed. Some rendering engines are too slow to render a large dataset quickly. In this patent, a list of possible views is generated. For each possible view includes a viewing position and viewing angle. For each possible view, rendering of a 3D dataset is performed to generate a corresponding preemptively rendered image for said each possible view. Each possible view and corresponding preemptively rendered image are stored. A head display unit's position and orientation are tracked and at a time epoch when said head display unit's position and orientation correspond to a possible view in said list of possible views, the corresponding preemptively rendered image is displayed.

Abstract: This patent provides a novel stereoscopic imaging system. In the preferred embodiment, the improved stereoscopic imaging system would be incorporated onto a smart phone, which is called the stereoscopic smart phone (SSP). The SSP would work in conjunction with one or more stereoscopic head display units (SHDUs). Once this system is operational, it will allow significant improvements to obtaining stereoscopic imagery. One novel aspect of this patent comprises wherein the stereoscopic cameras on the SSP can move in position to alter stereo separation distance and change the convergence to optimally image a scene.

Abstract: A method for detecting an object in an image includes: obtaining an image to be detected; generating a plurality of feature maps based on the image to be detected by a plurality of feature extracting networks in a neural network model trained for object detection, in which the plurality of feature extracting networks are connected sequentially, and input data of a latter feature extracting network in the plurality of feature extracting networks is based on output data and input data of a previous feature extracting network; and generating an object detection result based on the plurality of feature maps by an object detecting network in the neural network model.

Abstract: A generalizable and interpretable deep learning model for predicting biomarker status and biomarker metrics from histopathology slide images is provided.

Abstract: Provided are a method and a system for analyzing image data obtained by photographing a bridge by a drone using artificial intelligence, rapidly and accurately finding a part that requires maintenance of the bridge, and calculating a maintenance solution and a maintenance estimate for the part. The system for maintaining a bridge by analyzing bridge image data received from a drone using artificial intelligence, includes: the drone that photographs the bridge to generate the bridge image data; and an artificial intelligence bridge maintenance apparatus that finds a part of the bridge requiring maintenance, and calculates an optimal maintenance solution and an optimal maintenance estimate necessary for the bridge maintenance.

Abstract: An information processing apparatus according to an embodiment includes one or more hardware processors. The hardware processors obtain a first categorical distribution sequence corresponding to first input data and obtain a second categorical distribution sequence corresponding to second input data neighboring the first input data, by using a prediction model outputting a categorical distribution sequence representing a sequence of L categorical distributions for a single input data piece, where, L is a natural number of two or more. The hardware processors calculate, for each i of 1 to L, an inter-distribution distance between i-th categorical distributions in the first and second categorical distribution sequences. The hardware processors calculate a sum of L inter-distribution distances. The hardware processors update the prediction model's parameters to lessen the sum.

Abstract: A method is indicated for determining a confidence value of an object of a class detected in an input image with the aid of a trained neural network, including: producing an activation signature for the class of the detected object using a plurality of output images of a layer of the neural network, the input image being provided to the input of the neural network; scaling the activation signature to the dimension of the input image; comparing an object portion of the scaled activation signature with an activation signature distribution of all objects of the same class of a training data set of the neural network in order to determine the confidence value.

Abstract: A method for dynamically quantizing feature maps of a received image. The method includes convolving an image based on a predicted maximum value, a predicted minimum value, trained kernel weights and the image data. The input data is quantized based on the predicted minimum value and predicted maximum value. The output of the convolution is computed into an accumulator and re-quantized. The re-quantized value is output to an external memory. The predicted min value and the predicted max value are computed based on the previous max values and min values with a weighted average or a pre-determined formula. Initial min value and max value are computed based on known quantization methods and utilized for initializing the predicted min value and predicted max value in the quantization process.

Abstract: This invention provides a method to optimize an x-ray beam for more than one structure within the field of view. The preferred embodiment comprises a modular construction of a collimator comprising multiple materials of varying thickness. A first attenuation is performed by the first portion of the collimator to optimize a first anatomic feature and a second attenuation is performed by the second portion of the collimator to optimize a second anatomic feature.

Abstract: In this patent, an improved multi-user extended reality viewing technique is disclosed. A first user and a second user can be geographically separated and view the same volume. The first user can manipulate a virtual object and the second user can see the manipulated virtual object. A set of techniques are disclosed herein to cause the virtual object as it is presented to each user to be eye appealing and prevent dizziness and nausea.

Abstract: This patent provides a method and apparatus for acquiring imagery with improve spatial resolution through an apparatus called a “light painting imaging device”. Other aspects of this invention correct for barrel distortion and pincushion distortion.

Abstract: A system for training a computational neural network to recognise objects and/or actions from images, the system comprising: a training unit, comprising: an input interface configured to receive: a plurality of images captured from one or more cameras, each image having an associated timestamp indicating the time the image was captured; a data stream containing a plurality of timestamps each associated with an object and/or action; the data stream being generated by a system in an operative field of view of the one or more cameras; an image identification unit configured to identify from the plurality of images a set of images that each have a timestamp that correlates to a timestamp associated with an object and/or action from the data stream; a data-labelling unit configured to determine, for each image of the set of images, an image label that indicates the probability the image depicts: (i) an object of each of a set of one or more specified object classes; and/or (ii) a specified human action in dependen

Abstract: The disclosure relates to a network structure processing method and device and a related product, including: obtaining, by a computer device, an optimization instruction; and executing a corresponding optimization processing operation on the network structure according to the optimization level in the optimization instruction to obtain an optimized network structure. According to the processing method of the network structure, the resource cost can be reduced, and the detection rate of the network structure on images is improved.