Score Map Peak Position Detection Acceleration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting peak positions in score maps, such as those used in convolutional neural networks, are inefficient, particularly when dealing with large score maps or multiple peak points, and require significant processing time or pre-defined similar images, making them unsuitable for real-time object detection and machine learning applications.
Innovation Solution
An information processing apparatus and method that selects a position of interest in a score map by comparing its score with a threshold and determining if it indicates a peak position based on a relationship with scores within a predetermined region, thereby accelerating peak position detection by skipping unnecessary processing steps and reducing the number of positions to evaluate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If existing methods (PTL 1 or NPL 1) are used to detect peak positions in score maps, then peak position detection can be performed, but processing time is excessive and real-time detection is not achieved
Solution Approach 1:
The score map is divided into multiple blocks, and peak position detection is performed independently for each block. This segmentation allows parallel processing and reduces the overall processing time compared to analyzing the entire score map sequentially.
Solution Approach 2:
The patent performs partial peak position detection by focusing only on specific blocks within the score map rather than analyzing every position. This selective approach reduces processing time while still achieving accurate object detection by identifying peak positions in the most relevant regions.
2Measurement precision
If all positions in the score map are evaluated to ensure accurate peak detection, then detection accuracy is maintained, but processing complexity and time increase significantly
Solution Approach 1:
By dividing the score map into blocks and performing peak detection independently in each block, the patent reduces processing complexity while maintaining accuracy. Each block can be processed with simpler logic, and the results are combined to achieve overall accurate peak position detection.
Solution Approach 2:
The patent performs preliminary evaluation of blocks to identify which blocks contain potential peak positions before conducting detailed analysis. This preliminary filtering reduces the number of positions that require full evaluation, thereby reducing processing complexity while maintaining detection accuracy.
Data Source
AI summary
An information processing apparatus is provided. The apparatus determines, in order to detect a peak position in a score map, whether a position of interest in the score map indicates the peak position. The apparatus selects a new position of interest based on a relationship between (i) a score at the position of interest and (ii) a score at a first position within a region of a predetermined size around the position of interest.