Region of Interest Inferencing for Edge Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection techniques require intensive processing power and memory resources, particularly for high-resolution images, which is a challenge for low-power edge devices like battery-powered cameras without compromising detection accuracy.
Innovation Solution
The implementation of a computer vision system that maps an input image into grid cells, prioritizes regions of interest (ROI) for inferencing using a compact object detection model, and updates priority values based on inferencing results to optimize resource usage and predict future ROIs, allowing for reduced computational load and accurate object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection is performed on high-resolution images using existing techniques, then detection accuracy is improved, but processing power and memory resources are excessively consumed
Solution Approach 1:
The patent divides the input image into multiple grid cells and processes only selected regions of interest rather than the entire image. This segmentation approach reduces the computational domain from the full high-resolution image to smaller manageable regions, significantly lowering processing power requirements while maintaining detection accuracy through focused analysis of relevant areas.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on their importance. Regions containing detected objects or potential objects of interest receive more intensive processing, while other regions receive reduced or no processing. This local quality differentiation optimizes resource allocation by concentrating computational efforts where they are most needed.
2Measurement precision
If object detection is performed on high-resolution images using existing techniques, then detection accuracy is improved, but memory resources are excessively consumed
Solution Approach 1:
The patent segments the image processing task into multiple grid cells, allowing memory to be allocated and released in smaller units rather than requiring the entire high-resolution image to remain in memory simultaneously. This reduces peak memory usage while maintaining the ability to process high-resolution data through sequential or selective access to different regions.
Solution Approach 2:
The patent extracts and processes only the relevant portions of the image (regions of interest) rather than loading and processing the entire high-resolution image. This extraction approach reduces memory requirements by eliminating the need to store and manipulate unnecessary image data while preserving detection accuracy for the extracted regions.
3Reliability
If the entire input image is processed for object detection, then comprehensive detection coverage is achieved, but processing time and resource consumption increase
Solution Approach 1:
The patent performs preliminary processing steps such as grid cell division and initial region identification before conducting full object detection. This preliminary action enables the system to identify and prioritize regions of interest in advance, allowing subsequent detection to focus only on relevant areas rather than processing the entire image, thus improving processing efficiency while maintaining detection coverage.
Solution Approach 2:
The patent applies partial processing to the image by selectively analyzing only certain grid cells or regions rather than performing exhaustive processing on the entire image. This partial action approach maintains adequate detection coverage for regions of interest while reducing overall processing time and resource consumption by omitting or reducing processing in less critical areas.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This disclosure provides methods, devices, and systems for object detection. The present implementations more specifically relate to region of interest (ROI) inferencing techniques that can be implemented using a single object detection model. In some aspects, a computer vision system maps a set of grid cells to an input image so that each grid cell includes a respective portion of the image, and where each of the grid cells is assigned a respective priority value. The system selects an ROI of the image based on the priority value assigned to each grid cell and performs, on the ROI, an inferencing operation associated with an object detection model. The system updates the priority values for one or more of the grid cells based on a result of the inferencing operation. The system then selects another ROI based on the updated priority values and perform the inferencing operation on the new ROI.