Abstract: A method and system for processing images for a search is provided, including: receiving a plurality of images selected from search results; for each image in the plurality of images, retrieving a feature vector associated with the image; selecting a subset of the feature vectors based on similarity of feature vectors associated with the images in the plurality of images; and performing a search for feature vectors in a database similar to the feature vectors in the subset of feature vectors.

Abstract: A method for displaying video streams involves detecting multiple video events, and displaying multiple video streams on a display. Display windows for displayed video streams define an ordered sequence of display windows. When a new event is detected, the display of one or more of the video streams is shifted to respective next one or more display windows in an ordered sequence of display windows.

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: Methods, systems, and techniques for tagless tracking of an object-of-interest are disclosed. Image and non-image data are generated across a plurality of camera-specific regions, and the object-of-interest is tracked over a global map formed as a composite of these regions.

Abstract: A method of detecting an anomaly is provided, including dividing each frame of a video stream into a plurality of cells; in each cell formulate statistics based on metadata generated for the frame, the metadata related to presence of an object in the cell, velocity of objects in the cell, direction of motion of objects in the cell, and classification of objects in the cell; and using the formulated statistics to determine when the anomalous activity has occurred in one of the cells of the plurality of cells.

Abstract: Methods, systems, and techniques for enhancing use of two-dimensional (2D) video analytics by using depth data. Two-dimensional image data representing an image comprising a first object is obtained, as well as depth data of a portion of the image that includes the first object. The depth data indicates a depth of the first object. An initial 2D classification of the portion of the image is generated using the 2D image data without using the depth data. The initial 2D classification is stored as an approved 2D classification when the initial 2D classification is determined consistent with the depth data. Additionally or alternatively, a confidence level of the initial 2D classification may be adjusted depending on whether the initial 2D classification is determined to be consistent with the depth data, or the depth data may be used with the 2D image data for classification.

Abstract: Methods, systems, and techniques for monitoring an object-of-interest within a region involve receiving at least data from two sources monitoring a region and correlating that data to determine that an object-of-interest depicted or represented in data from one of the sources is the same object-of-interest depicted or represented in data from the other source. Metadata identifying that the object-of-interest from the two sources is the same object-of-interest is then stored for later use in, for example, object tracking.

Abstract: Multiple natural language training text strings are obtained. For example, text portions may be randomly selected and converted into natural language text based on one or more randomly selected rules. A formatted training text string is generated for each natural language training text string, for example using a context-free grammar parser. The formatted training text strings are inputted to a machine learning model. For each formatted training text string, using the machine learning model, a natural language text string is generated. The natural language text string is associated with one of the natural language training text strings. One or more parameters of the machine learning model are adjusted based on one or more differences between at least one of the natural language text strings and its associated natural language training text string.

Abstract: A camera includes a housing, a light source positioned within the housing, and a light-refracting apparatus. The light-refracting apparatus comprises a collimator shaped to collimate light emitted by the light source, and a lens comprising an at least partially concave light-emitting surface positioned to receive light collimated by the collimator and shaped to disperse the collimated light. The light-refracting apparatus is arranged to cause the dispersed light to be transmitted from within the housing into a field of view region of the camera.

Abstract: Compensating for delay in a Pan-Tilt-Zoom (PTZ) camera system is disclosed. Client-side view transformation is carried out to emulate a future Field Of View (FOV) of the camera so that the impact of latency is reduced.

Abstract: A system and method for playing back video feeds includes determining from the video feeds a subset of currently available video feeds, determining a playback layout based on the number of currently available video feeds and playing back the subset of currently available video feeds within the playback layout. The system and method further includes determining at least one change in availability in a given video feed of the plurality of video feeds, updating the subset of currently available video feeds based on the at least one change, updating the playback layout based on the updated subset of currently available video feeds and playing back the updated subset of currently available video feeds within the updated playback layout.

Abstract: A system for streaming multiple encodings of a source video stream is disclosed. A source video device system captures and/or receives source video comprising a plurality of source video frames. The source video device system encodes the source video frames into virtual frames, with each of the virtual frames being encoded using at least one different encoding parameter. The source video device system forms a container frame from the virtual frames and transmits the container frame over a network. In an example embodiment, the container frame comprises the virtual frames associated with a particular source frame and a container timestamp applicable to all of the virtual frames in the container frame.

Abstract: Methods, systems, and techniques for tracing callstacks, the results of which can be used to identify a source of a memory leak. A memory heap is sampled at different times to determine counts of respective allocations of different allocations at the different times. From the counts of the respective allocations at the different times, rates at which the respective allocations have changed over the different times, and which of the rates is equal to or greater than a leak indication rate, are determined. Callstacks for respective allocations of the different allocation sizes for each of the rates that is equal to or greater than the leak indication rate are traced. The results of the tracing may be used during debugging to identify the source of the memory leak.

Abstract: A computer implemented method for managing data received from a video source device is disclosed. At least a first storage tier and a second storage tier on a storage medium are identified for respectively storing different categories of data. As data streams are received from the video source, the data streams are associated with at least the first storage tier or the second storage tier and data from the received streams are stored in the associated storage tier. When the storage medium is full or meets some other specified maximum capacity threshold, data is deleted beginning with data that has an age that exceeds one or more specified maximum retention times, followed by data with the oldest age from one or both of the first and second storage tiers.

Abstract: A multi-headed camera assembly includes a base comprising a plurality of tracks, a plurality of image sensor assemblies each comprising a support mechanism configured to allow a quick and efficient assembly of a lens assembly within a housing and each removably coupled to the base by a corresponding support frame among a plurality of support frames. Each support frame comprises a guide member for guiding tilt movement of a corresponding image sensor assembly and a support base movably coupled to a corresponding track so as to accomplish pan movement of the corresponding image sensor assembly. The multi-headed camera assembly further includes a surface mount mechanism, an in-ceiling mount mechanism, or a pendant mount mechanism so that the multi-headed camera assembly can be removably mounted to a surface, mounted into a ceiling, or connected to a pendant base.

Abstract: A method for displaying video streams involves detecting multiple video events, and displaying multiple video streams on a display. Display windows for displayed video streams define an ordered sequence of display windows. When a new event is detected, the display of one or more of the video streams is shifted to respective next one or more display windows in an ordered sequence of display windows.

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: Alias capture to support searching for an object-of-interest is disclosed. A method includes capturing, using a camera with a defined field of view, video image frames that include a moving object-of-interest. The method also includes tracking the object-of-interest over a period of time starting when the object-of-interest enters the field of view and ending when the object-of-interest exits the field of view. The method also includes detecting, at a point in time in-between the start and end of the period of time of the tracking, a threshold exceeding change in an appearance of the object-of-interest. The method also includes creating, before the end of the period of time of the tracking, a new object profile for the object-of-interest in response to the detecting of the threshold exceeding change.

Abstract: A method, system and apparatus for image capture, analysis and transmission are provided. A link aggregation method involves identifying controller network ports to a source connected to the same subnetwork; producing packets associating corresponding controller network ports selected by the source CPU for substantially uniform selection; and transmitting the packets to their corresponding network ports. An image analysis method involves producing by a camera an indication whether a region of an image differs by a threshold extent from a corresponding region of a reference image; transmitting the indication and image data to a controller via a communications network; and storing at the controller the image data and the indication in association therewith. The controller may perform operations according to positive indications.

Abstract: Methods, systems, and techniques for classifying and/or detecting objects using visible and invisible light images. A visible light image and an invisible light image are received at a convolutional neural network (CNN). The visible light image depicts a region-of-interest imaged using visible light. The invisible light image depicts at least a portion of the region-of-interest imaged using invisible light, and at least one of the images depicts an object-of-interest within the portion of the region-of-interest shared between the images. The CNN then classifies and/or detects the object-of-interest using the images. The CNN may be trained to perform this classification and/or detection using pairs of visible and invisible light training images.