Neural Network Image Filtering for IoT Workload Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Environment control systems face high workload issues due to the need to process image flows at high frame rates, which can prevent other applications from running concurrently, especially in IoT devices with limited processing capabilities, and not all images in the flow are necessary for detection tasks like person presence or movement analysis.
Innovation Solution
A computing device equipped with a neural network that determines whether to process images based on a predictive model, reducing the frame rate by only processing relevant images, thereby reducing the workload and allowing other applications to run parallelly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the image flow is processed at a high frame rate (e.g., 30 FPS) by a neural network, then the detection accuracy and responsiveness are improved, but the processor workload becomes excessively high, preventing other applications from running concurrently
Solution Approach 1:
The patent segments the image processing task into two stages: first, a lightweight neural network filters images to identify only those containing persons; second, a more comprehensive image processing function is applied only to the filtered subset. This segmentation reduces overall processor workload while maintaining detection accuracy for applications that require it.
Solution Approach 2:
The patent applies partial processing to all images (the lightweight filtering step) and excessive/detailed processing only to necessary images (those containing persons). This partial/excessive action approach optimizes resource allocation by avoiding full processing of irrelevant images while ensuring thorough processing when needed.
2Loss of information
If all images in the image flow are processed by the image processing module, then complete information is available for all detection tasks, but the processing time and energy consumption increase significantly
Solution Approach 1:
The patent performs preliminary filtering using a lightweight neural network before the main image processing step. This preliminary action identifies and flags images containing persons, allowing the system to skip detailed processing of images without persons, thereby reducing processing time while maintaining information completeness for relevant images.
Solution Approach 2:
The patent extracts and processes only the necessary subset of images (those containing persons) from the complete image flow. By taking out and processing only relevant images with the full image processing module, the system maintains information completeness for detection tasks while significantly reducing overall processing time and energy consumption.
Data Source
AI summary
Method and computing device using a neural network to determine whether or not to process images of an image flow. A predictive model of the neural network is generated and stored at a computing device. The computing device receives (b) an image of the image flow and executes (c) the neural network, using the predictive model for generating an indication of whether or not to process the image based on input(s) of the neural network, the input(s) comprising the image. The computing device determines (d) whether or not to process the image by an image processing module, based on the indication of whether or not to process the image. The image is processed by the image processing module if the determination is positive and not processed if the determination is negative. Steps (b), (c), (d) are repeated for consecutive images of the image flow.