A feature extractor and novel training objective are provided for content-based image retrieval. For example, a computer-implemented method includes applying a query image and a search image to a neural network of a feature extraction network of a computing device, the query image indicating an object to be searched for in the search image. The feature extraction network includes the neural network, a spatial feature neural network receiving a first output of the neural network pertaining to the search image, and an embedding network receiving a second output of the neural network pertaining to the query image. The method includes generating spatial search features from the spatial feature neural network, generating a query feature from the embedding network, applying the query feature to an artificial neural network (ANN) index, and determining an optimal matching result of an object in the search image based on an operation using the ANN index.

System and methods enable vehicles to make ethical/empathetic driving decisions by using deep learning aided location intelligence. The systems and methods identify moral islands/complex driving scenarios where a complex ethical decision is required. A Generative Adversarial Network (GAN) is used to generate synthetic training data to capture varied ethically complex driving situations. Embodiments train a deep learning model (ETHNET) that is configured to output one or more driving decisions to be taken when a vehicle comes across an ethically complex driving situations in the real world.

Locating, aiding, and communicating with users and personnel in emergency situations by traversing a defined path utilizing an unmanned vehicle, detecting a user within a threshold distance of the defined path, logging a geolocation of the user within the unmanned vehicle, and determining whether to dispatch assistance to the user.

Systems and methods for identifying inclined regions are provided. In one aspect, a method is provided that includes receiving shadow data for at least one first ground object in a first region, wherein each first ground object is depicted in one overhead image of the first region, wherein the shadow data comprises a length of the respective first ground object as identified from the respective overhead image; receiving shadow data for at least one second comparable ground object in a second region, wherein each second ground object is depicted in one overhead image of the second region, wherein the shadow data comprises a length of the respective second ground object as identified from the respective overhead image; calculating a statistical measure describing the variability of the shadow lengths between objects in the first region and the second region; comparing the statistical measure to a predetermined threshold; and based on the comparison, identifying that the first region is inclined relative to the second region.

The present invention relates to a method for size estimation by image recognition of a specific target using a given scale. First, a reference objected is recognized in an image and the corresponding scale is established. Then the specific target is searched and the size of the specific target is estimated according to the acquired scale.

Aspects of the present disclosure relate to systems and methods for active depth sensing. An example apparatus configured to perform active depth sensing includes a projector. The projector is configured to emit a first distribution of light during a first time and emit a second distribution of light different from the first distribution of light during a second time. A set of final depth values of one or more objects in a scene is based on one or more reflections of the first distribution of light and one or more reflections of the second distribution of light. The projector may include a laser array, and the apparatus may be configured to switch between a first plurality of lasers of the laser array to emit light during the first time and a second plurality of laser to emit light during the second time.

Aspects of the present disclosure relate to systems and methods for active depth sensing. An example apparatus configured to perform active depth sensing includes a projector. The projector is configured to emit a first distribution of light during a first time and emit a second distribution of light different from the first distribution of light during a second time. A set of final depth values of one or more objects in a scene is based on one or more reflections of the first distribution of light and one or more reflections of the second distribution of light. The projector may include a laser array, and the apparatus may be configured to switch between a first plurality of lasers of the laser array to emit light during the first time and a second plurality of laser to emit light during the second time.

The present disclosure provides an earthquake prediction method and system based on ground-air remote sensing coupling. The method includes: acquiring a geomagnetic resonance cell; determining an initial earthquake magnitude, an epicentral distance, and an eruption time based on the geomagnetic resonance cell; determining an epicenter based on the epicentral distance; obtaining a satellite remote sensing cloud image and/or an infrared remote sensing image; determining an initial earthquake magnitude, an epicenter and an earthquake eruption time based on the satellite remote sensing cloud image and/or the infrared remote sensing image; and determining a final earthquake magnitude, a final epicenter and a final earthquake eruption time by analysis by a coupled system based on the geomagnetic resonance cell, the satellite remote sensing cloud image and/or the infrared remote sensing image. By the above method, an earthquake can be predicted.

The present disclosure provides an earthquake prediction method and system based on ground-air remote sensing coupling. The method includes: acquiring a geomagnetic resonance cell; determining an initial earthquake magnitude, an epicentral distance, and an eruption time based on the geomagnetic resonance cell; determining an epicenter based on the epicentral distance; obtaining a satellite remote sensing cloud image and/or an infrared remote sensing image; determining an initial earthquake magnitude, an epicenter and an earthquake eruption time based on the satellite remote sensing cloud image and/or the infrared remote sensing image; and determining a final earthquake magnitude, a final epicenter and a final earthquake eruption time by analysis by a coupled system based on the geomagnetic resonance cell, the satellite remote sensing cloud image and/or the infrared remote sensing image. By the above method, an earthquake can be predicted.

A hazard visualization system that can use artificial intelligence to identify locations at which hazards have occurred and a cause therein and to predict locations at which hazards may occur in the future is described herein. As a result, the hazard visualization system may reduce the likelihood of structural damage and/or loss of life that could otherwise occur due to natural disasters or other hazards. For example, the hazard visualization system can train an artificial intelligence model to predict the date, time, type, severity, path, and/or other conditions of a hazard that may occur at a geographic location. As another example, the hazard visualization system can train an artificial intelligence model to identify equipment or other infrastructure depicted in geographic images.