Abstract: A method includes obtaining training data for a classifier, the training data comprises one or more target classes, obtaining candidate background classes, selecting negative classes from the candidate background classes, wherein the negative classes exclude candidate background classes that are close to the target classes, wherein the negative classes exclude candidate background classes that are very different from the target classes, and wherein the negative classes include candidate background classes that are similar to the target classes, and training the classifier on a combined set of the selected negative classes and target classes.

Abstract: A video monitoring and analysis system detect subjects when they are entering and/or exiting from a room. The system enables a user to define a portal, such as doorway of the room. The system then monitors the movement of foreground objects in the room. Objects that appear only in the portal are classified as passing by the portal, e.g., doorway. Objects that initially appear in the portal and then are detected moving within the room are classified as having entered the room. Objects that are in the room and then disappear within the portal are classified as having exited the room. The system further has provisions for generating real-time alerts and performing forensic searches.

Abstract: Methods, systems, and computer programs are presented for training a deep neural network (DNN). One method includes an operation for training a predecessor network defined for image recognition of items, where parameters of a predecessor classifier are initialized with random numbers sampled from a predetermined distribution, and the predecessor classifier utilizes an image-classification probability function without bias. The method further includes an operation for training a successor network defined for image recognition of items in a plurality of classes, where parameters of a successor classifier are initialized with parameters learned from the predecessor network, and the successor classifier utilizes the image-classification probability function without bias. Further, the method includes operations for receiving an image for recognition, and recognizing the image utilizing the successor classifier.

Abstract: Hyperlapse results are generated from wide-angled, panoramic video. A set of wide-angled, panoramic video data is obtained. Video stabilization is performed on the obtained set of wide-angled, panoramic video data. Without user intervention, a smoothed camera path is automatically determined using at least one region of interest that is determined using saliency detection and semantically segmented frames of stabilized video data resulting from the video stabilization. A set of frames is determined to vary the velocity of wide-angled, panoramic rendered display of the hyperlapse results.

Abstract: Methods, systems, and computer programs are presented for training a deep neural network (DNN). One method includes an operation for training a predecessor network defined for image recognition of items, where parameters of a predecessor classifier are initialized with random numbers sampled from a predetermined distribution, and the predecessor classifier utilizes an image-classification probability function without bias. The method further includes an operation for training a successor network defined for image recognition of items in a plurality of classes, where parameters of a successor classifier are initialized with parameters learned from the predecessor network, and the successor classifier utilizes the image-classification probability function without bias. Further, the method includes operations for receiving an image for recognition, and recognizing the image utilizing the successor classifier.

Abstract: Methods, systems, and computer programs are presented for training a deep neural network (DNN). One method includes an operation for training a predecessor network defined for image recognition of items, where parameters of a predecessor classifier are initialized with random numbers sampled from a predetermined distribution, and the predecessor classifier utilizes an image-classification probability function without bias. The method further includes an operation for training a successor network defined for image recognition of items in a plurality of classes, where parameters of a successor classifier are initialized with parameters learned from the predecessor network, and the successor classifier utilizes the image-classification probability function without bias. Further, the method includes operations for receiving an image for recognition, and recognizing the image utilizing the successor classifier.

Abstract: Methods, systems, and computer programs are presented for training a deep neural network (DNN). One method includes an operation for training a predecessor network defined for image recognition of items, where parameters of a predecessor classifier are initialized with random numbers sampled from a predetermined distribution, and the predecessor classifier utilizes an image-classification probability function without bias. The method further includes an operation for training a successor network defined for image recognition of items in a plurality of classes, where parameters of a successor classifier are initialized with parameters learned from the predecessor network, and the successor classifier utilizes the image-classification probability function without bias. Further, the method includes operations for receiving an image for recognition, and recognizing the image utilizing the successor classifier.

Abstract: A method includes obtaining training data for a classifier, the training data comprises one or more target classes, obtaining candidate background classes, selecting negative classes from the candidate background classes, wherein the negative classes exclude candidate background classes that are close to the target classes, wherein the negative classes exclude candidate background classes that are very different from the target classes, and wherein the negative classes include candidate background classes that are similar to the target classes, and training the classifier on a combined set of the selected negative classes and target classes.

Abstract: Hyperlapse results are generated from wide-angled, panoramic video. A set of wide-angled, panoramic video data is obtained. Video stabilization is performed on the obtained set of wide-angled, panoramic video data. Without user intervention, a smoothed camera path is automatically determined using at least one region of interest that is determined using saliency detection and semantically segmented frames of stabilized video data resulting from the video stabilization. A set of frames is determined to vary the velocity of wide-angled, panoramic rendered display of the hyperlapse results.

Abstract: An integrated surveillance system combining video surveillance and data from other sensor-based security networks is used to identify activities that may require attention.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: An integrated surveillance system combining video surveillance and data from other sensor-based security networks is used to identify activities that may require attention.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: An integrated surveillance system combining video surveillance and data from other sensor-based security networks is used to identify activities that may require attention.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: An integrated surveillance system combining video surveillance and data from other sensor-based security networks is used to identify activities that may require attention.

Abstract: Via intuitive interactions with a user, robots may be trained to perform tasks such as visually detecting and identifying physical objects and/or manipulating objects. In some embodiments, training is facilitated by the robot's simulation of task-execution using augmented-reality techniques.

Abstract: An integrated surveillance system combining video surveillance and data from other sensor-based security networks is used to identify activities that may require attention.