Adjusting parameters in a network-connected security system based on content analysis
Systems and methods are described for adjusting the parameters in a network-connected security system based on analysis of content generated by electronic devices in the network-connected security system. In an example embodiment, content such as video captured by a video surveillance camera is processed to analyze the performance of the network-connected security system. Based on the processing, updated parameters are selected to configure and improve the performance of the network-connected security system.
1. A method comprising:
receiving, by a computer system, content generated by an electronic device of a security system;
detecting, by the computer system, one or more physical objects captured in the content using computer-vision techniques;
processing, by the computer system, the content using machine learning, the processing comprising classifying the one or more physical objects based on one or more criteria;
determining, by the computer system, a parameter for the electronic device based on the processing, such that a performance metric of the security system is increased when the electronic device is configured using the parameter;
configuring, by the computer system, the electronic device using the parameter; and
exchanging signaling information with a user device to determine a path between the computer system and the user device based on locations of the computer system and the user device for sending the content to the user device via the path using a peer-to-peer (P2P) connection.
2. The method of claim 1 , wherein the parameter is any one of an optical parameter, an image processing parameter, or an encoding parameter.
3. The method of claim 1 , wherein the content comprises video files.
4. The method of claim 1 , wherein classifying the one or more physical objects comprises identifying, by the computer system, instances of different classes of the one or more physical objects.
5. The method of claim 1 , wherein classifying the one or more physical objects comprises deep learning-based video recognition using a plurality of representational layers.
6. The method of claim 1 , wherein classifying the one or more physical objects comprises distinguishing between instances of the one or more physical objects using at least one of Regions with Convolutional Neural Network Features (RCNN), Fast RCNN, Single Shot Detector (SSD), or You Only Look Once (YOLO).
7. The method of claim 1 , further comprising determining an identity of the detected physical objects using machine-learning appearance-based models.
8. The method of claim 1 , wherein processing the content comprises training a machine-learning model using labeled image data.
9. The method of claim 1 , further comprising:
accessing, by the computer system, historical user feedback data based on previous content generated by the electronic device, wherein:
the performance metric of the security system is based at least on the historical user feedback data.
10. A computer configured to:
receive content generated by an electronic device of a security system;
detect one or more physical objects captured in the content using computer-vision techniques;
process the content using machine learning, the processing comprising classifying the one or more physical objects based on one or more criteria;
determine a parameter for the electronic device based on the processing, such that a performance metric of the security system is increased when the electronic device is configured using the parameter;
configure the electronic device using the parameter; and
exchange signaling information with a user device to determine a path between the computer system and the user device based on locations of the computer system and the user device for sending the content to the user device via the path using a peer-to-peer (P2P) connection.
11. The computer of claim 10 , wherein the electronic device is a network-connected video camera.
12. The computer of claim 10 , wherein the parameter is any one of an optical parameter, an image processing parameter, or an encoding parameter.
13. The computer of claim 10 , wherein the content comprises video files.
14. The computer of claim 10 , configured to classify the one or more physical objects by identifying instances of different classes of the one or more physical objects.
15. The computer of claim 10 , configured to classify the one or more physical objects by deep learning-based video recognition using a plurality of representational layers.
16. The computer of claim 10 , configured to classify the one or more physical objects by distinguishing between instances of the detected physical objects using at least one of Regions with Convolutional Neural Network Features (RCNN), Fast RCNN, Single Shot Detector (SSD), or You Only Look Once (YOLO).
17. The computer of claim 10 , configured to classify the one or more physical objects by determining an identity of the detected physical objects using machine-learning appearance-based models.
18. The computer of claim 10 , further configured to process the content by training a machine-learning model using labeled image data.
19. A computer-readable non-transitory storage medium storing computer instructions, which when executed by one or more computer processors cause the one or more computer processors to:
receive content generated by an electronic device of a security system;
detect one or more physical objects captured in the content using computer-vision techniques;
process the content using machine learning, the processing comprising classifying the one or more physical objects based on one or more criteria;
determine a parameter for the electronic device based on the processing, such that a performance metric of the security system is increased when the electronic device is configured using the parameter;
configure the electronic device using the parameter; and
exchange signaling information with a user device to determine a path between the computer system and the user device based on locations of the computer system and the user device for sending the content to the user device via the path using a peer-to-peer (P2P) connection.