Object Position Circuit Reducing AI Load via Partial Region Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face identification systems using deep learning or neural networks are computationally intensive, leading to overloading of the AI module when processing large image data, which increases design and manufacturing costs.
Innovation Solution
An object position determination circuit that detects the position of an object in a partial region of subsequent frames based on the detection result of previous frames, reducing the computational load by only processing specific regions within each frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning or neural networks are utilized to analyze and process an image to identify the position of a human face, then the object position detection accuracy is improved, but the computation amount increases and the AI module becomes overloaded
Solution Approach 1:
The patent divides the image into multiple regions of interest (ROIs) based on object detection results from previous frames. Instead of processing the entire image with deep learning models, the system segments the image and only processes relevant regions in subsequent frames, reducing computational load while maintaining detection accuracy.
Solution Approach 2:
The patent performs object detection on previous frames to determine the positions of objects beforehand. These preliminary detection results are then used to define regions of interest for subsequent frames, allowing the AI module to focus computation only on relevant areas rather than processing the entire image each time.
2Productivity
If the artificial intelligence module has greater ability to handle large image data, then the object position detection capability is improved, but the design and manufacturing costs increase
Solution Approach 1:
The patent segments the image processing task into two parts: a lightweight object detection module that operates on previous frames to identify object positions, and a deep learning module that only processes the segmented regions of interest in subsequent frames. This segmentation allows the system to achieve high detection capability without requiring the entire AI module to be oversized, thereby reducing design and manufacturing costs.
Solution Approach 2:
The patent applies partial action by only processing the necessary portions of the image (regions of interest) rather than the entire image. This partial processing approach enables the system to achieve sufficient detection capability without over-provisioning the AI module, thus avoiding increased design and manufacturing costs.
3Reliability
If the entire image is processed in each frame, then the object position detection completeness is improved, but the processing time increases
Solution Approach 1:
The patent segments the image into regions of interest based on object detection results from previous frames. By only processing these segmented regions in subsequent frames rather than the entire image, the system maintains detection completeness for objects of interest while significantly reducing processing time.
Solution Approach 2:
The patent performs preliminary object detection on previous frames to identify object positions and define regions of interest beforehand. This preliminary action allows the system to skip unnecessary processing areas in subsequent frames, maintaining detection completeness while reducing processing time through targeted processing.
Data Source
AI summary
The present invention provides an object position determination circuit including a receiving circuit and a detecting circuit. In operations of the object position determination circuit, the receiving circuit receives an image signal; and the detecting circuit detects a position of an object in an Nth frame of the image signal, determines a partial region within an (N+M)th frame of the image signal according to the position of the object in the Nth frame, and only detects the partial region within the (N+M)th frame to determine a position of the object in the (N+M)th frame, wherein N and M are positive integers.