Abstract: Embodiments for training a deep neural network by applying tags to enhanced versions of a raw data set are disclosed. In one embodiment, the raw data set may comprise, for example, images, and the method may comprise receiving the raw data set of images, applying an enhancement to the images, deriving tags based on the enhanced images while maintaining a correlation to the original raw images, and using the raw images and the tags to train a deep neural network.

Abstract: In a method of processing video images, a processor might obtain a video stream, select a set of processing pipelines from a plurality of processing pipelines, apply the set of processing pipelines to the video stream to obtain a data set of analytics data, wherein the pipeline analytics data for a processing pipeline of the set of processing pipelines comprises data about images of the video images, store the pipeline analytics data of each processing pipeline into a queryable data structure, receive queries from a user interface, and return portions of the video stream, or references thereto, in response to the queries.

Abstract: In a method of processing video images, a processor might obtain a video stream, select a set of processing pipelines from a plurality of processing pipelines, apply the set of processing pipelines to the video stream to obtain a data set of analytics data, wherein the pipeline analytics data for a processing pipeline of the set of processing pipelines comprises data about images of the video images, store the pipeline analytics data of each processing pipeline into a queryable data structure, receive queries from a user interface, and return portions of the video stream, or references thereto, in response to the queries.

Abstract: A data-enhanced video viewing system scans videos in order to detect and extract certain objects, such as human faces, so as to compile non-time based synopses, including “facelines” of people appearing in the video sequence. It can also provide a time-based synopsis that includes timestamps for all detected objects. The data-enhanced video viewing system can be deployed on a network for a client to request data extraction on one or more designated videos. The designated video may be one of those that have been uploaded to social networks, uploaded to online video hosting sites, streamed over the Internet or other network, or uploaded directly by the user.

Abstract: A Convolutional Neural Network (CNN) includes an initial set of convolutional layers and max pooling units, in which any input is convoluted with the learned image filters and the output is a stack of the different filter responses. Max pooling produces a scaled version of the output. The process can be repeated several times, resulting in a stack of space invariant-scaled images. Since the operation is space invariant, the computations of these layers not need to be recomputed if interested just in certain regions of the image. A Region Of Interest (ROI) Pooling layer is used to select regions to be processed by the set of fully connected layers, which uses the response of the multiple convolutional layers of the network to determine the regions where the objects (of different scales) could be located. This object proposal method is implemented as a Region Of Interest (ROI) Selector.

Abstract: A computer vision pipeline detects tracks and classifies people or other specified class of objects in a steam of video. The ability to not only detect motion, but to distinguish people or other specified objects, can improve the systems usefulness in applications like security monitoring. A motion detection module provides a motion bitmap and a background subtraction module provides a foreground bitmap, and an object tracking module uses these bitmaps identify and track the specified classes of objects. From these objects and tracks, categorized object data can then be generated.

Abstract: A Convolutional Neural Network (CNN) includes an initial set of convolutional layers and max pooling units, in which any input is convoluted with the learned image filters and the output is a stack of the different filter responses. Max pooling produces a scaled version of the output. The process can be repeated several times, resulting in a stack of space invariant-scaled images. Since the operation is space invariant, the computations of these layers not need to be recomputed if interested just in certain regions of the image. A Region Of Interest (ROI) Pooling layer is used to select regions to be processed by the set of fully connected layers, which uses the response of the multiple convolutional layers of the network to determine the regions where the objects (of different scales) could be located. This object proposal method is implemented as a Region Of Interest (ROI) Selector.

Abstract: A data-enhanced video viewing system scans videos in order to detect and extract certain objects, such as human faces, so as to compile non-time based synopses, including “facelines” of people appearing in the video sequence. It can also provide a time-based synopsis that includes timestamps for all detected objects. The data-enhanced video viewing system can be deployed on a network for a client to request data extraction on one or more designated videos. The designated video may be one of those that have been uploaded to social networks, uploaded to online video hosting sites, streamed over the Internet or other network, or uploaded directly by the user.

Abstract: A computer vision pipeline detects tracks and classifies people or other specified class of objects in a steam of video. The ability to not only detect motion, but to distinguish people or other specified objects, can improve the systems usefulness in applications like security monitoring. A motion detection module provides a motion bitmap and a background subtraction module provides a foreground bitmap, and an object tracking module uses these bitmaps identify and track the specified classes of objects. From these objects and tracks, categorized object data can then be generated.