Dual Inference Engine Video Event Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
AI systems require large amounts of GPU and CPU processor power to process large amounts of data for reliable event detection in video streams, leading to resource-intensive bandwidth and processor load, which can be costly and impractical for continuous monitoring.
Innovation Solution
The method involves receiving a time slice of images, storing a first set of images at a lower frame rate, and using two inference engines with trained machine learning models to detect trigger and detail events at different frame rates, reducing unnecessary processing and resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI systems process large amounts of video data at high frame rates for reliable event detection, then detection reliability is improved, but processor power consumption and resource load increase substantially
Solution Approach 1:
The patent segments the video processing task into two distinct stages: a first inference stage that processes video data at a lower frame rate to detect coarse events, and a second inference stage that processes selected video data at a higher frame rate to detect detailed events. This segmentation allows the system to maintain detection reliability while reducing overall processor power consumption by avoiding continuous high-rate processing of all video data.
Solution Approach 2:
The patent implements dynamic processing rates by adjusting the frame rate at which video data is processed based on the detection stage and content characteristics. The system dynamically switches between a first processing rate for initial event detection and a second processing rate for detailed analysis, optimizing the balance between detection reliability and energy consumption rather than using a static high processing rate throughout.
2Measurement precision
If AI systems continuously process video streams at high frame rates, then event detection accuracy is improved, but bandwidth and computational costs increase exponentially
Solution Approach 1:
The patent applies partial action by processing only a subset of video data at the higher second processing rate. Specifically, the system processes video data at the first lower rate for all frames, then selectively processes only certain frames or regions at the second higher rate based on detected events or regions of interest. This approach maintains detection accuracy for critical events while significantly reducing the total volume of data processed.
Solution Approach 2:
The patent segments video data processing into two categories: data processed at the first lower processing rate for general monitoring, and data processed at the second higher processing rate for detailed analysis. This segmentation of processing volumes allows the system to maintain high detection accuracy for important events while reducing overall bandwidth and computational costs by limiting high-rate processing to necessary portions of the video stream.
3Reliability
If the system processes all video frames at the main frame rate, then no events are missed, but processing time and resource consumption increase
Solution Approach 1:
The patent applies preliminary action by first processing video data at the lower first processing rate to identify potential events or regions of interest before conducting more detailed processing at the higher second processing rate. This preliminary screening allows the system to maintain event detection completeness by catching all potential events at the lower rate, then investing additional processing time only when necessary for detailed analysis, rather than uniformly processing all frames at the higher rate.
Data Source
AI summary
The invention provides event detection method for detecting a detail event in a time sequence of images having a main frame rate, the method comprising:receiving a time slice of the time sequence of images;storing in a memory a first set of images from the time slice at a first frame rate which is equal to or lower than the main frame rate;providing a first inference engine comprising a first trained machine learning model which is trained for detecting a trigger event in an input comprising at least one image;providing a second inference engine comprising a second trained machine learning model which is trained for detecting the detail event in an input comprising at least one image;processing a second set of images from the time slice at a second frame rate which is lower than the first frame rate, by providing at least one image of the second set of images as input to the first inference engine for detecting the trigger event in the second set of images;upon detection of the trigger event in the second set of images, processing the first set of images from the memory by providing at least one image of the first set of images as input to the second inference engine for detecting the detail event.