Event Detector Training for Low-Power Video Surveillance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-driven object detection solutions for real-time video surveillance are computationally intensive, requiring powerful GPUs, making them costly and prone to latency when run on remote servers or in the cloud, which limits accessibility and increases costs for consumers.
Innovation Solution
An event detection system that trains a specialized event detector using object and motion detectors to generate inter-frame differences, reducing processing power requirements and improving detection accuracy by learning temporal characteristics and attributing motion to specific objects, allowing for efficient detection of events like 'Person taking out the trash' or 'Person mowing' without continuous GPU access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning driven object detection solutions are used, then detection accuracy is improved, but computational cost and processing power requirements increase
Solution Approach 1:
The system segments the detection task into two parts: a deep learning-based object detector that runs offline to generate training data, and a simplified event detector that runs in real-time on resource-constrained devices. This segmentation allows the complex deep learning model to be used only during the offline training phase, while the real-time system uses lightweight algorithms that require minimal computational power.
Solution Approach 2:
The system performs preliminary action by using the deep learning object detector to generate labeled training data in advance. The offline system processes video frames to create annotated datasets showing object locations and motion patterns, which are then used to train the simplified event detector. This preliminary processing eliminates the need to run complex deep learning models during real-time surveillance.
2Power
If deep learning object detection is run on remote servers or cloud, then processing capability is improved, but latency in event notifications increases
Solution Approach 1:
The system extracts the event detection functionality from remote cloud servers and implements it locally on edge devices or embedded systems. By taking out the detection algorithm and deploying it directly on the surveillance device, the system eliminates network transmission delays and achieves real-time detection with immediate local notification capability.
3Measurement precision
If GPU based object detection algorithms are used, then detection accuracy is improved, but cost increases
Solution Approach 1:
The system replaces expensive, continuously-operating GPU hardware with a software-based solution that uses simplified detection algorithms. The offline deep learning model generates training data once, and the resulting lightweight event detector runs on inexpensive processors without requiring continuous GPU access. This substitutes costly hardware with a more economical software approach that achieves the same real-time detection goal.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an event detector. The methods, systems, and apparatus include actions of obtaining frames of a video, determining whether an object of interest is detected within the frames, determining whether motion is detected within the frames, determining whether the frames correspond to motion by an object of interest, generating a training set that includes labeled inter-frame differences based on whether the frames correspond to motion by an object of interest, and training an event detector using the training set.