Abstract: Computing a height of a building is performed by inputting a pair of two-dimensional (2-D) aerial images of a city along with their metadata. Using the metadata, a three-dimensional (3-D) vector from each image toward the location of the camera when each image was taken is determined. A plurality of pairs of corresponding image points from the images are computed, in each pair of image points an image point of one image identifies the same physical point on the building as the second image point of the second image. Next, the images are superimposed, and for each pair of image points, determine the intersection of the 3-D vector of the first image originating at the first image point with the 3-D vector of the second image originating at the second image point. Each intersection is a 3-D position and the height is determined from the median of these 3-D positions.

Abstract: The present principles relates to a method for intra-predictive encoding a coding unit comprising picture data, said intra-predictive encoding depending on a prediction tree and a transform tree, characterized in that the method further comprises: —obtaining said prediction tree by spatially partitioning the coding unit according to a non-square partition type; —determining said transform tree from said coding unit in order that each of its leaves is embedded into a unique unit of said obtained prediction tree; and —signaling in a signal the size of the leaves of said transform tree and said a non-square partition type.

Abstract: Techniques for building and managing data models are provided. According to certain aspects, systems and methods may enable a user to input parameters associated with building one or more data models, including parameters associated with sampling, binning, and other factors. The systems and methods may automatically generate program code that corresponds to the inputted parameters and display the program code for review by the user. The systems and methods may build the data models and generate charts and plots depicting aspects of the data models. Additionally, the systems and methods may combine data models and select champion data models.

Abstract: In some embodiments, the present invention provides for an exemplary inventive convolutional neural network-based and computer-implemented method for identifying and evaluating pizza, including: collecting video input representative of the food and received from at least one camera, applying a first CNN to select, from the video input, a set of best pizza containing video frames of a particular pizza from the plurality of pizza containing video frames; applying the first CNN to identify a best pizza containing image from the set, to localize at least one pizza portion of the particular pizza in the identified best pizza containing image, and to determine a type of the pizza of the particular pizza from the identified best pizza containing image; applying a second CNN to determine a map of pizza components of the particular pizza by automatically performing pizza image segmentation and to automatically score the particular pizza based on the determined map of pizza components.

Abstract: The present disclosure provides a method for visual question answering, which relates to a field of computer vision and natural language processing. The method includes: acquiring an input image and an input question; constructing a Visual Graph based on the input image, wherein the Visual Graph comprises a Node Feature and an Edge Feature; updating the Node Feature by using the Node Feature and the Edge Feature to obtain an updated Visual Graph; determining a question feature based on the input question; fusing the updated Visual Graph and the question feature to obtain a fused feature; and generating a predicted answer for the input image and the input question based on the fused feature. The present disclosure further provides an apparatus for visual question answering, a computer device and a non-transitory computer-readable storage medium.

Abstract: A system for distinguishing a type of excreta deposited in a toilet is disclosed. The system includes a toilet and a processor. The toilet has a bowl adapted to receive multiple types of excreta from a user and a sensor which monitors the volume of excreta deposited in the toilet. The processor compares excreta volume data derived from the sensor to a database comprising excreta-type volume data and associates a time segment from the excreta volume data as representing an excreta-type. This system can provide data which may be used to determine the rate of excreta deposit into the toilet and associated those rates with excreta events types such as urination or defecation.

Abstract: An apparatus of labeling for object detection according to an embodiment of the present disclosure includes an image selector that determines a plurality of labeling target images from among a plurality of unlabeled images, and determines a labeling order of the plurality of labeling target images, a feedback obtainer that obtains label inspection information on the plurality of labeling target images from a user, and a model trainer that learns the label inspection information input from the user by using the labeling target images, obtains a pseudo label for supervised learning based on a learning result using the label inspection information, and re-determines the labeling order of the labeling target images based on the pseudo label.

Abstract: A binary logic circuit for performing an interpolation calculation between two endpoint values E0 and E1 using a weighting index i for generating an interpolated result P, the values E0 and E1 being formed from Adaptive Scalable Texture Compression (ASTC) low-dynamic range (LDR) colour endpoint values C0 and C1 respectively, the circuit comprising: an interpolation unit configured to perform an interpolation between the colour endpoint values C0 and C1 using the weighting index i to generate a first intermediate interpolated result C2; and combinational logic circuitry configured to receive the interpolated result C2 and to perform one or more logical processing operations to calculate the interpolated result P according to the equation P=?((C2<<8)+C2+32)/64? when the interpolated result is not to be compatible with an sRGB colour space, and according to the equation P=?((C2<<8)+128·64+32)/64? when the interpolated result is to be compatible with an sRGB colour space.

Abstract: A binary logic circuit for performing an interpolation calculation between two endpoint values E0 and E1 using a weighting index i for generating an interpolated result P, the values E0 and E1 being formed from Adaptive Scalable Texture Compression (ASTC) colour endpoint values C0 and C1 respectively, the colour endpoint values C0 and C1 being low-dynamic range (LDR) or high dynamic range (HDR) values, the circuit comprising: an interpolation unit configured to perform an interpolation between the colour endpoint values C0 and C1 using the weighting index i to generate a first intermediate interpolated result C2; combinational logic circuitry configured to receive the interpolated result C2 and to perform one or more logical processing operations to calculate the interpolated result P according to the equation: (1) P=?(C2<<8)+C2+32)/64? when the interpolated result is not to be compatible with an sRGB colour space and the colour endpoint values are LDR values; (2) P=?(C2<<8)+128.

Abstract: A data processing device for motion segmentation in images obtained by cameras that move in a background environment includes an input for receiving a temporal sequence of images from the cameras and a processor. The processor is adapted for, for at least two images, of the temporal sequence of images, that are obtained by at least two cameras at different points in time, determining epipoles, defining corresponding image regions of limited image disparity due to parallax around the epipoles in the at least two images, and applying a motion segmentation algorithm to the corresponding image regions. Warping is applied to the corresponding image regions to compensate for camera rotation and misalignment beyond a threshold value.

Abstract: An apparatus and method for successive multi-frame image denoising are herein disclosed. The apparatus includes a first subtractor including a first input to receive a frame of the image, a second input to receive a reference frame, and an output; an absolute value function block including an input connected to the output of the first subtractor and an output; a second subtractor including a first input connected to the output of the absolute value function block, a second input for receiving a first predetermined value, and an output; and a maximum value divider function block including an input connected to the output of the second subtractor and an output for outputting filter weights.

Abstract: Pairing a user response and associated context with a neural network associated with a virtual assistant computer during a dynamic text conversation with an end user. The virtual assistant computer receives a detected user generated text input; determines context of the detected user generated text input; compares the context of the detected user generated text input by comparing a confidence score representing context of the user generated input to a classification associated with each of a plurality of existing nodes of a neural network. For confidence scores below a threshold relative to the classification associated with each of the existing nodes of the neural network, the virtual assistant computer creates a new node within the neural network and assigns the context of the user generated text to the new node.

Abstract: A computer implemented method of generating a gradient boosting decision tree for obtaining predictions includes finding split points by sorting variable values of a feature by their gradient during training of the gradient boosting decision tree, performing a linear search to find a subset of variables with maximum split gain, and modifying a node of the gradient boosting decision tree to have multiple split points on the node for a feature as a function of the linear search. In a further example, a computer implemented method of controlling overfitting in a gradient boosting decision tree includes combining values of low population feature values into a virtual bin, fanning out the virtual bin into feature values having a low population, and including the low population feature values into multiple split points on a node of the gradient boosting decision tree.

Abstract: An approach is provided for determining an optimal alignment of a device. The approach, for example, involves receiving image data from a device mounted in a vehicle. The approach also involves presenting an alignment template in a user interface of the device as an overlay on the image data, wherein the alignment template provides one or more guidelines indicating a target alignment of the device to capture images from the vehicle for an application. The approach further involves processing the image data against the alignment template to determining an alignment of the device in relation to the target alignment.

Abstract: The disclosure provides an image recognition method and an image recognition device. The method includes: acquiring an image and capturing a plurality of feature points in the image; obtaining a capsule network, where the capsule network sequentially includes a convolution layer, a primary capsule layer, a routing capsule layer, and an output layer; inputting the image and the feature points into the convolution layer to generate a plurality of feature vectors; inputting the feature vectors and the feature points into the primary capsule layer to generate a plurality of activity vectors; and generating a recognition result corresponding to the image by the routing capsule layer and the output layer based on the activity vectors.

Abstract: Methods and systems for outputting depth data during a medical procedure on a patient. Depth data is outputted, representing at least one of relative depth data and general depth data. Tracking information about the position and orientation of a medical instrument and depth information about variations in depth over a site of interest are used. Relative depth data represents the depth information relative to the position and orientation of the instrument. General depth data represents the depth information over the site of interest independently of the position and orientation of the instrument.

Abstract: A method for generating a neural network for detecting one or more objects in images includes generating one or more self-supervised proposal learning losses based on the one or more proposal features and corresponding proposal feature predictions. One or more consistency-based proposal learning losses are generated based on noisy proposal feature predictions and the corresponding proposal predictions without noise. A combined loss is generated using the one or more self-supervised proposal learning losses and one or more consistency-based proposal learning losses. The neural network is updated based on the combined loss.

Abstract: A system identifies anomalies in an image of an object. An input image of the object containing zero or more anomalies is supplied to an image encoder. The image encoder generates an image model. The image model is applied to an image decoder that forms a substitute non-anomalous image of the object. Differences between the input image and the substitute non-anomalous image identify zero or more areas of the input image that contain the zero or more the anomalies. The system implements a flow-based model and has been trained using (a) a set of augmented anomaly-free images of the object applied at the image encoder and (b) a reconstruction loss calculated based on a norm of differences between each augmented anomaly-free image of the object and a corresponding output image from the image decoder.

Abstract: An image compression apparatus includes: an image acquisition unit configured to acquire a raw data image; a pre-processing network configured to receive the raw data image and pre-process the raw data image according to a pattern estimation method learned beforehand; and an encoder unit configured to receive the pre-processed image and compress the pre-processed image according to a pre-designated standard compression technique to output a compressed image. The pre-processing network, which can be added during learning and can be implemented as an artificial neural network, can have learned beforehand by way of a backpropagation of a restoration error through a codec modeling unit that has learned beforehand to simulate a standard codec unit, where the restoration error can be obtained by comparing a restored image obtained based on a simulated decoded image with the raw data image.

Abstract: A method for automatically analyzing and constructing communications to a plurality of recipients includes automatically separating communication content files into page groups in a system comprising one or more intelligent communication design servers, wherein each of the page groups is associated a recipient of the communications, inputting the communication content files into an intra-page machine prediction model to produce intra-page parameters, inputting the communication content files and the intra-page parameters into an intra-page machine prediction model to produce intra-group parameters and inter-group parameters, automatically constructing standard communication design files by an intelligent communication content learning and constructing engine based on the communication content files and the intra-page parameters, intra-group parameters, and inter-group parameters, and printing and finishing physical mailing pieces to be mailed to the recipients based on the standard communication design files.