A computer acquires training data, in which first image data, object information indicating first objects included in the first image data, and relationship information indicating a first relationship between the first objects are associated. The computer executes machine learning that trains, based on the training data, an inference model that infers both second objects included in second image data and a second relationship between the second objects or selectively infers one of the second objects and the second relationship according to an input of the second image data to the inference model. The machine learning uses a penalty term when calculating an error between an inference result of the inference model and the training data. The penalty term causes the error to increase as an overlap between inferred image regions, which are inferred to be image regions in which objects are present in the inference result, increases.

The invention relates to a system and method for monitoring a structural vulnerability not yet apparent as a defect. A type of vulnerability is identified from a grouping of elements and/or features of a structure and its systems, called findings, whose attributes meet a particular set of selection rules called relationship-association classifiers (RACs)—a series of one or more tests of relationships between attributes of findings, the series of tests indicating whether findings are associated and collectively classified as a particular type of vulnerability or defect. A RAC may be established by an optimization function or a neural network function, trained from maintenance reports and documented findings and attributes concurrent with each maintenance report. Findings are identified and attributes evaluated from one or more images, typically taken during routine maintenance of the structure. A grouping identified as a vulnerability can be reported to a maintenance entity for follow-up.

In various examples, systems and methods are disclosed that preserve rich spatial information from an input resolution of a machine learning model to regress on lines in an input image. The machine learning model may be trained to predict, in deployment, distances for each pixel of the input image at an input resolution to a line pixel determined to correspond to a line in the input image. The machine learning model may further be trained to predict angles and label classes of the line. An embedding algorithm may be used to train the machine learning model to predict clusters of line pixels that each correspond to a respective line in the input image. In deployment, the predictions of the machine learning model may be used as an aid for understanding the surrounding environment—e.g., for updating a world model—in a variety of autonomous machine applications.

A method for classifying and counting objects recoverable from a urine sample processed onto a slide. The method includes the following steps: receiving at least one digitalized image of the whole slide; detecting connected components by segmentation of the image of the whole slide; classifying the detected connected components into countable connected components and uncountable connected components using a classifier; for the countable connected components using an object detection model to obtaining the number of objects for each class; for the uncountable components using a semantic segmentation model to obtaining the number of objects for each class; summing up the number of objects for each class obtained from the semantic segmentation model and the object detection model outputting a number of objects for each class.

Various embodiments relate to a computer apparatus for the bone microstructure connectivity recovery of a skeletal image reconstructed through an artificial neural network using the representations of a node-link graph-based bone microstructure and a method thereof. The computer apparatus and the method may be configured to represent a node-link graph from a bone microstructure of an input skeletal image, reinforce a connectivity of the bone microstructure in the node-link graph, and change the node-link graph into a skeletal image.

Computer-readable media, methods, and systems are disclosed for converting a diagram into a digital diagram format. A diagram is received as an image file and a plurality of recognition stages produce a final diagram based on predicted information from one or more of the recognition stages. The plurality of recognition stages including a shape detection stage for detecting a plurality of shapes and at least one arrow detection stage in which arrows are detected as relations between pairs of shapes. The final diagram is generated based on the predicted information and is converted into the digital diagram format compatible with a diagram modeling language.

Augmented reality (AR) systems, devices, media, and methods are described for generating AR experiences including interactions with virtual or physical prop objects. The AR experiences are generated by capturing images of a scene with a camera system, identifying an object receiving surface and corresponding surface coordinates within the scene, identifying an AR primary object and a prop object (physical or virtual), establishing a logical connection between the AR primary object and the prop object, generating AR overlays including actions associated with the AR primary object responsive to commands received via a user input system that position the AR primary object adjacent the object receiving surface responsive to the primary object coordinates and the surface coordinates within the scene and that position the AR primary object and the prop object with respect to one another in accordance with the logical connection, and presenting the generated AR overlays with a display system.

The present invention discloses a system and a method for providing multi-channel high-quality depth estimation from a monocular camera for providing augmented reality (AR) and virtual reality (VR) features to an image. The invention further includes the method to enhance generalization on deployment-friendly monocular depth inference pipeline with semantic information. Furthermore, a vivid and intact reconstruction is guaranteed by inpainting the missing depth and context within the single image input.

Techniques are described for automated operations involving capturing and analyzing information from an interior of a house, building or other structure, for use in generating and providing a representation of that interior. Such techniques may include using a user's mobile device to capture visual data from multiple viewing locations (e.g., video captured while the mobile device is rotated for some or all of a full 360 degree rotation at each viewing location) within multiple rooms, capturing data linking the multiple viewing locations, analyzing each viewing location's visual data to create a panorama image from that viewing location, analyzing the linking data to determine relative positions/directions between at least some viewing locations, creating inter-panorama links in the panoramas to each of one or more other panoramas based on such determined positions/directions, and providing information to display multiple linked panorama images to represent the interior.

Computer-readable media, methods, and systems are disclosed for converting a diagram into a digital diagram format. A diagram is received as an image file and a plurality of recognition stages produce a final diagram based on predicted information from one or more of the recognition stages. The plurality of recognition stages including a shape detection stage for detecting a plurality of shapes and at least one arrow detection stage in which arrows are detected as relations between pairs of shapes. The final diagram is generated based on the predicted information and is converted into the digital diagram format compatible with a diagram modeling language.