Object Position Circuit Reducing AI Compute Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face identification systems using deep learning or neural networks require high computation, leading to increased design and manufacturing costs due to the need for more powerful AI modules to handle large image data effectively.
Innovation Solution
An object position determination circuit that identifies object positions in partial image frames and estimates positions in other frames through extrapolation, reducing the computational load on the AI module by performing object position detection less frequently and using prediction methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning or neural networks are utilized to analyze and process image data to identify object positions, then identification accuracy is improved, but computation amount increases and AI module complexity increases
Solution Approach 1:
The image processing task is segmented into two parts: (1) traditional image processing circuit handles basic frame processing and displays all frames, (2) AI module only processes selected key frames (e.g., every Mth frame) to detect object positions. This segmentation reduces the computational burden on the AI module while maintaining identification accuracy through selective application of deep learning.
Solution Approach 2:
The AI module performs object position detection periodically rather than continuously - specifically on every Mth frame (e.g., every 2nd, 3rd, or 5th frame). The traditional image processing circuit handles intermediate frames without AI processing. This periodic action significantly reduces computation amount while maintaining effective monitoring coverage.
2Measurement precision
If AI module processes all image frames to ensure accurate object position detection, then detection accuracy is maintained, but processing speed decreases and energy consumption increases
Solution Approach 1:
Processing tasks are segmented between two circuits: the traditional image processing circuit processes all frames at high speed for display, while the AI module processes only selected key frames for accurate object position detection. This segmentation enables both high overall processing speed and accurate detection where needed.
Solution Approach 2:
The AI module operates periodically on every Mth frame rather than continuously on all frames. This periodic operation maintains detection accuracy on key frames while achieving high overall processing throughput, as the traditional circuit handles intermediate frames without AI computation overhead.
3Adaptability or versatility
If a more powerful AI module is used to handle larger image data, then processing capability is improved, but design and manufacturing costs increase
Solution Approach 1:
The system segments processing responsibilities so that the AI module only handles object position detection on selected frames, while a traditional image processing circuit handles other frame processing tasks. This segmentation allows the use of a less powerful, lower-cost AI module while maintaining overall system capability to handle large image data effectively.
Solution Approach 2:
The AI module performs detection on a partial subset of frames (every Mth frame) rather than all frames. This partial action approach provides sufficient processing capability for accurate object tracking while avoiding the need for an overly powerful and expensive AI module that would be required to process every frame with the same level of analysis.
Data Source
AI summary
The present invention provides an object position determination circuit including a receiving circuit, a detecting circuit and a calculating circuit. In the operations of the object position determination circuit, the receiving circuit is configured to receive an Nth frame and an (N+M)th frame of an image signal, where N is a positive integer, and M is a positive integer greater than one; the detecting circuit is configured to detect positions of an object in the Nth frame and the (N+M)th frame; and the calculating circuit is configured to generate a position of the object in an (N+M+A)th frame according to the positions of the object in the Nth frame and the (N+M)th frame, wherein A is a positive integer.