Abstract: Methods and systems are provided for automatic classification of information in an adversarial machine learning setting. For example, a learner that includes multiple classifiers may perform data classification automatically. Each classifier may be trained with adversarial data of a different strength. For a given query to be classified, the learner is configured to intelligently select a classifier that is commensurate with the adversarial strength of the data inside the query, without explicit knowledge of that data or its adversarial strength.

Abstract: A computer implemented method for processing a seismic image comprising seismic values obtained from seismic measurements performed on a geological formation includes determining a seismic dip image based on the seismic image, said seismic dip image comprising local seismic dips representative of the local gradient of the seismic values of the seismic image; initializing a seismic horizon surface modeled by using a combination of spline functions or by using a triangle mesh; and iteratively modifying coefficients used for combining the spline functions, thereby iteratively modifying the seismic horizon surface to progressively increase alignment between local orientations of the seismic horizon surface and the corresponding local seismic dips, until a predetermined stop criterion is satisfied.

Abstract: A method for detecting and locating a lesion in a medical image is provided. A target medical image of a lesion is obtained and input into a deep learning model to obtain a target sequence. A first feature map output from the last convolution layer in the deep learning model is extracted. A weight value of each network unit corresponding to each preset lesion type in a fully connected layer is extracted. For each preset lesion type, a fusion feature map is calculated according to the first feature map and the corresponding weight value and resampled to the size of the target medical image to generate a generic activation map. The maximum connected area in each generic activation map is determined, and a mark border surrounding the maximum connected area is created. A mark border corresponding to each preset lesion type is added to the target medical image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a storage device, for electric grid modeling using surfel data are enclosed. An electric grid wire identification method includes: obtaining a set of surface elements (surfels), wherein each surfel of the set of surfels represents a portion of a surface of an object in a geographic region; selecting, based on one or more surfel attributes, one or more surfels of the set of surfels that each represent a portion of a surface of an electric grid wire; generating a representation of the electric grid wire from the selected one or more surfels; and adding the representation of the electric grid wire to a virtual model of the electric grid. Obtaining the set of surfels can include obtaining ranging data of the geographic region; and generating the set of surfels from the ranging data.

Abstract: The present invention discloses an image coding/decoding method and an apparatus for the method. More specifically, a method for filtering an image by a decoding apparatus comprises deriving boundaries of a block divided into quad-tree plus binary-tree structure; determining an edge among boundaries of the block to which de-blocking filtering is applied; determining a type of de-blocking filtering to be applied to the edge; and performing de-blocking filtering to a picture sample restored according to the de-blocking filtering type, wherein the edge or the type of de-blocking filtering is determined by considering width or height of the block.

Abstract: This disclosure relates to ultrasound imaging with reduced speckle. Ultrasound imaging with frequency compounding and angle compounding is disclosed. Techniques are disclosed to make ultrasound imaging with frequency and angle compounding more robust. One such technique is non-rigid image registration to align ultrasound images for angle compounding. Another disclosed technique includes selecting a subset of ultrasound images for non-rigid ultrasound image registration.

Abstract: A method for analyzing one or more conditions of a transportation pathway includes obtaining, using an imaging device of an inspection system, image data reproducible as a plurality of images of the transportation pathway, each of the plurality of images being reproducible as an image of a portion of the transportation pathway, each portion of the transportation pathway having an associated location along a length of the transportation pathway, analyzing, using one or more processors of the inspection system, the image data to determine a first plurality of metrics indicative of a condition of the transportation pathway at each of the associated locations, and generating a first graph, using the determined first plurality of metrics, that is indicative of the condition of the transportation pathway at each of the associated locations.

Abstract: An unsupervised image segmentation method includes: performing a superpixel segmentation on an image containing a target object to acquire a plurality of superpixel sets, each superpixel set corresponding to a respective superpixel node; generating an undirected graph according to superpixel nodes; determining foreground superpixel nodes and background superpixel nodes in the undirected graph according to a first label set corresponding to the plurality of superpixel nodes; generating a minimization objective function according to the foreground superpixel nodes and the background superpixel nodes; segmenting the undirected graph according to the minimization objective function to acquire a foreground part and a background part and to generate a second label set; and performing an image segmentation on the image according to a comparison result of the first label set and the second label set.

Abstract: It is possible to estimate a slack level accurately in consideration of a shape of a deformed cable. A point cloud analysis device sets a plurality of regions of interest obtained by window-searching a wire model including a quadratic curve model representing a cable obtained from a point cloud consisting of three-dimensional points on an object, the region of interest being divided into a first region and a second region.

Abstract: Various implementations disclosed herein include devices, systems, and methods that generate floorplans and measurements using a three-dimensional (3D) representation of a physical environment generated based on sensor data.

Abstract: According to one embodiment, a reading system includes an extractor, a determiner, and a reader. The extractor extracts a candidate image from an input image. The candidate image is of a portion of the input image in which a segment display is imaged. The determiner calculates an angle with respect to a reference line of each of a plurality of straight lines detected from the candidate image, and determines whether or not the candidate image is an image of a segment display based on a distribution indicating a relationship between the angle and a number of the straight lines. The reader reads a numerical value displayed in a segment display from the candidate image determined to be an image of a segment display.

Abstract: A method includes receiving, from a camera disposed on a robotic device, a two-dimensional (2D) image of a body of an actor and determining, for each respective keypoint of a first subset of a plurality of keypoints, 2D coordinates of the respective keypoint within the 2D image. The plurality of keypoints represent body locations. Each respective keypoint of the first subset is visible in the 2D image. The method also includes determining a second subset of the plurality of keypoints. Each respective keypoint of the second subset is not visible in the 2D image. The method further includes determining, by way of a machine learning model, an extent of engagement of the actor with the robotic device based on (i) the 2D coordinates of keypoints of the first subset and (ii) for each respective keypoint of the second subset, an indicator that the respective keypoint is not visible.

Abstract: Comprehensive 2D learning images are collected for learning subjects. Standardized 2D gallery images of many gallery subjects are collected, one per gallery subject. A 2D query image of a query subject is collected, of arbitrary viewing aspect, illumination, etc. 3D learning models, 3D gallery models, and a 3D query model are determined from the learning, gallery, and query images. A transform is determined for the selected learning model and each gallery model that yields or approximates the query image. The transform is at least partly 3D, such as 3D illumination transfer or 3D orientation alignment. The transform is applied to each gallery model so that the transformed gallery models more closely resemble the query model. 2D transformed gallery images are produced from the transformed gallery models, and are compared against the 2D query image to identify whether the query subject is also any of the gallery subjects.

Abstract: A method and apparatus for detecting an object in a three-dimensional (3D) point cloud are disclosed. The apparatus includes a three-dimensional (3D) model reconstruction unit configured to reconstruct a three-dimensional (3D) model using three-dimensional (3D) point cloud data, a two-dimensional (2D) object detection unit configured to detect an object in at least one two-dimensional (2D) image captured to generate the three-dimensional (3D) point cloud data, a three-dimensional (3D) position estimation unit configured to form a cylinder surrounding the object in the three-dimensional (3D) model based on a two-dimensional (2D) position of the detected object so as to estimate a three-dimensional (3D) position of the detected object using the cylinder, and an object region extraction unit configured to acquire three-dimensional (3D) point cloud data contained in the cylinder so as to extract an object region from the acquired three-dimensional (3D) point cloud data.

Abstract: Subscriber identity modules can be bound to devices and utilized for authentication. An initial or additional subscriber identity module can be bound to a user device, for instance after successful user authentication by way of an alternate mechanism. A subscriber identity module of a device can be identified, and carrier data associated with the subscriber identity module can be requested and received. A determination can be made regarding whether the subscriber identity module is linked to the device, for instance by an original linkage. A user profile can be automatically updated with the carrier data to bind the user to the user device based on the subscriber identity module when the subscriber identity module is verified to be linked to the user device. Subsequently, the subscriber identity module can be utilized as a basis for authentication.

Abstract: In network devices, during manufacturing, input for designation of a region code to be a non-specific region code is stored in a BIOS memory of the network device, and a specific region code is stored off the BIOS. During boot up, the BIOS is checked for a specific region code to regulate wireless transmissions at a physical location of operation. Responsive to receiving the non-specific region code from BIOS, the specific region code is requested from a region code server based on a network device identifier. Once received, the region code is stored in flash memory, until rebooted or otherwise reset, rather than BIOS.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. The present implementations more specifically relate to selecting images that can be used for training a neural network to infer denoised representations of images captured in low light conditions. In some aspects, a machine learning system may obtain a series of images of a given scene, where each of the images is associated with a different SNR (representing a unique combination of exposure and gain settings). The machine learning system may identify a number of saturated pixels in each image and classify each of the images as a saturated image or a non-saturated image based on the number of saturated pixels. The machine learning system may then select the non-saturated image with the highest SNR as the ground truth image, and the non-saturated images with lower SNRs as the input images, to be used for training the neural network.

Abstract: Methods, systems, and apparatus for removing precipitation from video are disclosed. A method includes generating, from a first set of images of a scene from a camera, a segmented background image model of the scene; obtaining a second set of images from the camera; identifying, in an image of the second set of images, a plurality of edges, determining that a first edge of the plurality of edges satisfies criteria for representing precipitation based at least in part on determining that the first edge (i) does not correspond to the background image model of the scene and (ii) extends into two or more contiguous segments of the scene; in response, classifying each of the contiguous segments as a precipitation segment; generating pixel data for each of the precipitation segments; and applying the pixel data to each precipitation segment in the image.

Abstract: This document describes methods to assign a collective state to a group of traffic signal devices that are concurrently detected by multiple cameras. The states may be color states or other states. The system will process the images to identify states of each of the devices in the images. When the devices exhibit more than one state, the system will determine an overall state for the group by generating a confidence score for each of the states. The system will select, from the multiple states, the state having a confidence score that exceeds a threshold. The system will then use the selected state to assign an overall state to the group of traffic signal devices. The system may use the overall state to generate a signal that will cause the vehicle to perform an action such as a motion control action, or output of an audible or visual alert.

Abstract: A system includes a storage device storing a set instructions and a processor in communication with the storage device, wherein when executing the set of instructions, the processor is configured to cause the system to obtain raw data. The processor may also be configured to cause the system to determine one or more reconstruction-related algorithms and determine one or more containers for the one or more reconstruction-related algorithms. Each of the one or more containers may correspond to at least one of the one or more reconstruction-related algorithms. The system may also be configured determine a reconstruction flow based on the one or more containers and process the raw data according to the reconstruction flow to generate a target image.