Abstract: 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.

Abstract: A sequence of video frames of an area of interest is obtained. A first background model of the area of interest is constructed based on a first parameter. A second background model of the area of interest is constructed based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models is determined. A stationary target is determined based on the determined difference. An alert concerning the stationary target is generated.

Abstract: Techniques performed by a data processing system for operating a personalized oncology system herein include accessing a first histopathological image of a histopathological slide of a sample taken from a first patient; analyzing the first histopathological image using a first machine learning model configured to extract first features from the first histopathological image; searching a histological database that includes a plurality of second histopathological images and corresponding clinical data for a plurality of second patients to generate search results; analyzing the plurality of third histopathological images and the corresponding clinical data associated with the plurality of third histopathological images using statistical analysis techniques to generate associated statistics and metrics associated with mortality, morbidity, time-to-event, or a combination thereof for the plurality of third patients associated with the third histopathological images; and presenting an interactive visual representat

Abstract: A sequence of video frames of an area of interest is obtained. A first background model of the area of interest is constructed based on a first parameter. A second background model of the area of interest is constructed based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models is determined. A stationary target is determined based on the determined difference. An alert concerning the stationary target is generated.

Abstract: Techniques for alerting a human operator of a predicted side collision of a vehicle with a detected pedestrian or bicyclist herein including obtaining sensor data indicative of a trajectory of the vehicle, determining a predicted trajectory of the vehicle based on the sensor data, estimating a likelihood of collision between the detected pedestrian or bicyclist and the first side of the vehicle based on at least the first position of the pedestrian or bicyclist determined by the fusion module and the predicted trajectory of the vehicle, determining that a warning should be presented based on at least the estimated likelihood of collision, and presenting an alert to a human operator of the vehicle in response to the determination that the warning should be presented.

Abstract: Provided is a computer-implemented method for processing one or more video frames. The meth can include generating, by a processor, a change in value of one or more pixels obtained from the one or more video frames; classifying, by the processor, the change in value of the one or more pixels to produce one or more classes of the change in value of the one or more pixels, wherein the one or more classes include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing, by the processor, a listing of detected targets based on the one or more classes.

Abstract: Systems and methods for alerting a human operator of a predicted side collision of a transit bus with a pedestrian or bicyclist, including receiving range scanner data from a range scanner located at a first side of the bus covering a first area on the first side of the bus, receiving image data from a sensor located at the first side of the bus and covering an area along the first side of the bus that overlaps the first area, detecting and tracking the pedestrian or bicyclist based on the range scanner data and the image data, estimating a likelihood of collision between the pedestrian or bicyclist and the first side of the bus based on the tracking of the pedestrian or bicyclist, and presenting an alert to the human operator in response to a determination that a warning should be presented based on the estimated likelihood of collision.

Abstract: A collision warning system for vehicles includes: a detection, tracking, and localization (DTL) laser module receiving laser data from a first laser range scanner, and generating laser data output, wherein the first laser range scanner covers a first laser area; a detection, tracking, and localization (DTL) thermal module receiving thermal data from a first thermal video sensor, and generating thermal data output, wherein the first thermal video sensor covers a first thermal area; a fusion module receiving the laser data output and the thermal data output, fusing the laser data output and the thermal data output, and generating a situational awareness map; and a collision prediction module receiving the situational awareness map, predicting a collision between a detected object and a vehicle, and warning an operator regarding the predicted collision.

Abstract: Provided is a computer-implemented method for processing one or more video frames. The meth can include generating, by a processor, a change in value of one or more pixels obtained from the one or more video frames; classifying, by the processor, the change in value of the one or more pixels to produce one or more classes of the change in value of the one or more pixels, wherein the one or more classes include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing, by the processor, a listing of detected targets based on the one or more classes.

Abstract: Provided is a computer-implemented method for processing one or more video frames. The meth can include generating, by a processor, a change in value of one or more pixels obtained from the one or more video frames; classifying, by the processor, the change in value of the one or more pixels to produce one or more classes of the change in value of the one or more pixels, wherein the one or more classes include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing, by the processor, a listing of detected targets based on the one or more classes.

Abstract: Provided is a computer-implemented method for processing one or more video frames. The meth can include generating, by a processor, a change in value of one or more pixels obtained from the one or more video frames; classifying, by the processor, the change in value of the one or more pixels to produce one or more classes of the change in value of the one or more pixels, wherein the one or more classes include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing, by the processor, a listing of detected targets based on the one or more classes.

Abstract: A computer-implemented method for processing one or more video frames may include obtaining one or more video frames; generating one or more blobs using the one or more video frames; classifying the one or more blobs to produce one or more classified blobs, wherein the one or more classified blobs include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing a list of detected targets based on the one or more classified blobs.

Abstract: A computer-implemented method for processing one or more video frames may include obtaining one or more video frames; generating one or more blobs using the one or more video frames; classifying the one or more blobs to produce one or more classified blobs, wherein the one or more classified blobs include one or more of a stationary target, a moving target, a target insertion, a target removal, or a local change; and constructing a list of detected targets based on the one or more classified blobs.

Abstract: One or more video frames may be obtained, and a background model may be constructed based on a first parameter. A second background model may be constructed using the one or more video frames based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models may be determined. One or more stationary targets may be determined based on the determined difference. The one or more stationary targets may be classified. An alert concerning the one or more classified stationary targets may be generated.

Abstract: A sequence of video frames of an area of interest is obtained. A first background model of the area of interest is constructed based on a first parameter. A second background model of the area of interest is constructed based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models is determined. A stationary target is determined based on the determined difference. An alert concerning the stationary target is generated.

Abstract: A video imager may be used to measure the natural and/or artificial light levels in a space. The lighting estimates may be passed to a controller via a communication network. The lighting estimates may be in the form of actual radiance values, brightness, or other forms. The controller may determine if and which areas of the space require more light or if and which areas of the space have more light than required. Based on the lighting estimates, the controller may issue commands to turn on/off or dim/brighten the light from various ones or combinations of light sources in or around the space via actuators. The controller may also directly interface with the light sources.

Abstract: A video sequence of a field of view within an environment is received. Targets are detected in the video sequence. Target geo-positional information is received. Correspondences between the targets detected in the video sequence and the target geo-positional information are determined and used to calibrate the camera and to geo-register a field of view of the camera.

Abstract: A computer-based method for automatic detection of water regions in a video include the steps of estimating a water map of the video and outputting the water map to an output medium, such as a video analysis system. The method may further include the steps of training a water model from the water map; re-classifying the water map using the water model by detecting water pixels in the video; and refining the water map.

Abstract: A sequence of video frames of an area of interest is obtained. A first background model of the area of interest is constructed based on a first parameter. A second background model of the area of interest is constructed based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models is determined. A stationary target is determined based on the determined difference. An alert concerning the stationary target is generated.

Abstract: A computer-based method for automatic detection of water regions in a video include the steps of estimating a water map of the video and outputting the water map to an output medium, such as a video analysis system. The method may further include the steps of training a water model from the water map; re-classifying the water map using the water model by detecting water pixels in the video; and refining the water map.