Abstract: One aspect provides a method for object detection including detecting, using an electronic processor, a plurality of candidate objects in a video using a convolutional neural network detection process and a background subtraction detection process and identifying, using the electronic processor, a candidate object from the plurality of candidate objects. The candidate object detected by the background subtraction detection process in a location of the video with no candidate objects detected by the convolutional neural network detection process. The method also includes determining, using the electronic processor, a background subtraction confidence level of the candidate object and categorizing, using the electronic processor, the candidate object as a detected object in the video in response to the background subtraction confidence level satisfying a background subtraction confidence threshold.

Abstract: There is disclosed a method that includes analyzing, using an at least one processor, image data to generate or facilitate acquisition of violation assessment data associated with a vehicle or an owner of that associated vehicle. The method also includes receiving, at the at least one processor, input originating from a police officer indicating an intention to stop the vehicle. The method also includes making a determination, at the at least one processor, whether the violation assessment data supports affirming the stop of the vehicle as a compliant stop, or whether the violation assessment data supports disaffirming the stop of the vehicle. The method also includes transmitting a notification in respect of the stop of the vehicle to a display or speaker perceptible to the police officer. The notification informs as to compliancy or non-compliancy of the stop of the vehicle.

Abstract: A computer-implemented method of localizing an image of a person captured using a camera is provided, the person in the field of view of a camera, comprising: obtaining the image captured using a camera, the image comprising the person within a bounding box; determining at least one slant value associated with the person within the bounding box; determining head image coordinates and feet image coordinates for the person using the at least one slant value; and localizing the person by projecting the head image coordinates to a head plane and the feet image coordinates for the person to a ground plane.

Abstract: A method of detecting unusual motion is provided, including: determining features occurring during a fixed time period; grouping the features into first and second subsets of the fixed time period; grouping the features in each of the first and second subsets into at least one pattern interval; and determining when an unusual event has occurred using at least one of the pattern intervals.

Abstract: Clustering and active learning for teach-by-example, and methods therefor, are disclosed. One method includes clustering, at an at least one electronic processor: a plurality of first detections together as a first cluster based on each detection of the first detections corresponding to respective first image data being identified as potentially showing a first perceptible category of a plurality of perceptible categories; and a plurality of second detections together as a second cluster based on each detection of the second detections corresponding to respective second image data being identified as potentially showing a second perceptible category of the perceptible categories.

Abstract: A device, method and system for identifying objects in warped images from a fisheye camera is provided. A device converts warped images from a fisheye camera into dewarped images. The device determines that a first object identified in the dewarped images is: entering a blind-spot region of the dewarped images; or located at a first edge region of the dewarped images. The device determines that a second object is: exiting the blind-spot region of the dewarped images, or located at a second edge region of the dewarped images, the second edge region opposite the first edge region. The device, in response to the second object meeting one or more reidentification conditions associated with the first object, reidentifies the second object as the first object at the dewarped images.

Abstract: One example electronic computing device includes electronic processor configured to store a list of a plurality of video cameras being considered for an upgrade. The electronic processor may be further configured to receive first video footage captured by a first video camera and second video footage captured by a second video camera. The electronic processor may be further configured to analyze the first video footage and the second video footage to generate a plurality of upgrade priority scores. In response to determining that a first upgrade priority score of the first video camera is greater than a second upgrade priority score of the second video camera, the electronic processor may be further configured to output an upgrade priority notification configured to inform a user that the first video camera has a higher upgrade priority than the second video camera.

Abstract: Methods, systems, and techniques for automatic camera calibration. One or more calibration images are captured using a camera. The calibration images depict one or more bounding boxes, and each of the bounding boxes bounds a person. For each of the bounding boxes, the person is modeled using a rectangle or a parallelepiped, and a projection of the rectangle or parallelepiped is determined. A mapping that maps foot vertices of the projection to head vertices of the projection is determined, and using the foot vertices and the mapping, estimates of the head vertices and distances between the head vertices and the estimates of the head vertices are determined. The camera is calibrated by iteratively updating, using an objective function, the camera parameters so as to reduce those distances.

Abstract: A computer-implemented method of localizing an image of a person captured using a camera is provided, the person in the field of view of a camera, comprising: obtaining the image captured using a camera, the image comprising the person within a bounding box; determining at least one slant value associated with the person within the bounding box; determining head image coordinates and feet image coordinates for the person using the at least one slant value; and localizing the person by projecting the head image coordinates to a head plane and the feet image coordinates for the person to a ground plane.

Abstract: One embodiment provides an access control system including access control sensors to detect actions performed in a vicinity of an access point, a verification sensor to verify access of the access point, and an electronic processor communicatively coupled to the access control sensors and the verification sensor. The electronic processor is configured to in response to an access intent model satisfying an accuracy condition, deploy the access intent model for the access point and receive a dataset indicating an action performed in the vicinity of the access point. The electronic processor is also configured to predict an access intent to access the access point by applying the access intent model to the dataset and enable access through the access point. The electronic processor is further configured to receive verification data indicating whether the access point is accessed, and automatically assign a label to the dataset based on the verification data.

Abstract: There is provided an appearance search system comprising one or more cameras configured to capture video of a scene, the video having images of objects. The system comprises one or more processors and memory comprising computer program code stored on the memory and configured when executed by the one or more processors to cause the one or more processors to perform a method. The method comprises identifying one or more of the objects within the images of the objects. The method further comprises implementing a learning machine configured to generate signatures of the identified objects and generate a signature of an object of interest. The system further comprises a network configured to send the images of the objects from the camera to the one or more processors.

Abstract: A method, system and computer program product for self-learned and probabilistic-based prediction of inter-camera object movement is disclosed. The method includes building and storing a transition model defined by transition probability and transition time distribution data generated during operation of a first video camera and one or more other video cameras over time. The method also includes employing at least one balance flow algorithm on the transition probability and transition time distribution data to determine a subset of the video cameras to initiate a search for an object based on a query. The method also includes running the search for the object over the subset of the video cameras.

Abstract: A method of detecting unusual motion is provided, including: determining features occurring during a fixed time period; grouping the features into first and second subsets of the fixed time period; grouping the features in each of the first and second subsets into at least one pattern interval; and determining when an unusual event has occurred using at least one of the pattern intervals.

Abstract: Methods, systems, and techniques for automatic camera calibration. One or more calibration images are captured using a camera. The calibration images depict one or more bounding boxes, and each of the bounding boxes bounds a person. For each of the bounding boxes, the person is modeled using a rectangle or a parallelepiped, and a projection of the rectangle or parallelepiped is determined. A mapping that maps foot vertices of the projection to head vertices of the projection is determined, and using the foot vertices and the mapping, estimates of the head vertices and distances between the head vertices and the estimates of the head vertices are determined. The camera is calibrated by iteratively updating, using an objective function, the camera parameters so as to reduce those distances.

Abstract: A method of detecting unusual motion is provided, including: determining features occurring during a fixed time period; grouping the features into first and second subsets of the fixed time period; grouping the features in each of the first and second subsets into at least one pattern interval; and determining when an unusual event has occurred using at least one of the pattern intervals.

Abstract: There is provided an appearance search system comprising one or more cameras configured to capture video of a scene, the video having images of objects. The system comprises one or more processors and memory comprising computer program code stored on the memory and configured when executed by the one or more processors to cause the one or more processors to perform a method. The method comprises identifying one or more of the objects within the images of the objects. The method further comprises implementing a learning machine configured to generate signatures of the identified objects and generate a signature of an object of interest. The system further comprises a network configured to send the images of the objects from the camera to the one or more processors.

Abstract: There is provided an appearance search system comprising one or more cameras configured to capture video of a scene, the video having images of objects. The system comprises one or more processors and memory comprising computer program code stored on the memory and configured when executed by the one or more processors to cause the one or more processors to perform a method. The method comprises identifying one or more of the objects within the images of the objects. The method further comprises implementing a learning machine configured to generate signatures of the identified objects and generate a signature of an object of interest. The system further comprises a network configured to send the images of the objects from the camera to the one or more processors.

Abstract: A camera system comprises an image capturing device, and connected to it are an object classification module and a calibration module. The object classification module is operable to determine whether or not an object in an image is a member of an object class, and the calibration module is operable to estimate representative sizes of the object. The object classification module may determine a confidence parameter that is used by the calibration module, or conversely, the calibration module may produce a size that is used by the classification module.

Abstract: Methods, systems, and techniques for object detection and tracking are provided. A system may include a module configured to generate a plurality of region proposals, each region proposal comprising a part of a video frame, a CNN pre-trained for object detection, the plurality of region proposals being input to the CNN; a tracker for tracking one or more targets based on outputs from the CNN across the series of video frames and generating tracking information on the one or more targets; and a module further configured to refine the plurality of region proposals to be input to the CNN, based on the tracking information.

Abstract: Methods, systems, and techniques for classifying an object-of-interest using an artificial neural network, such as a convolutional neural network. An artificial neural network receives a sample image including the object-of-interest overlaying a background and a sample background image excluding the object-of-interest and corresponding to the background overlaid by the object-of-interest. The object-of-interest is classified using the artificial neural network. The artificial neural network classifies the object-of-interest using the sample background and sample images. Prior to receiving the sample background and sample images the artificial neural network has been trained to classify the object-of-interest using training image pairs. Each of at least some of the training image pairs includes a first training image that includes a training object-of-interest overlaying a training background and a training background image excluding the training object-of-interest and corresponding to the training background.