Deep Learning Localization Apparatus Reliability Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing localization methods using deep learning models struggle to accurately determine the reliability of inferred position coordinates, leading to potential incorrect operations when peak values in heatmaps are erroneous, as they conflate position coordinates with reliability.
Innovation Solution
A localization apparatus that includes a deep learning model trained with teacher data featuring pixel groups distinct from the subject, a position coordinate calculation unit to determine coordinates, and a reliability calculation unit to assess reliability using global information from the inference image data, allowing for separate evaluation of position coordinates and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the peak value of the heatmap is used to indicate both position coordinates and reliability, then the output is simple and easy to interpret, but the reliability becomes incorrect when the peak value is erroneous
Solution Approach 1:
The patent segments the heatmap output into two independent components: position coordinates (derived from peak location) and reliability (derived from peak width). This segmentation allows each component to be evaluated independently, resolving the contradiction where the peak value simultaneously represents both position and reliability, causing reliability to become incorrect when position is erroneous.
2Device complexity
If the same scale is used for position coordinates and reliability, then the output format is simplified, but the detection performance decreases due to conflation of position and reliability information
Solution Approach 1:
The patent changes the parameter scale used for reliability from the same scale as position coordinates to a different scale based on peak width. This parameter change allows reliability to be expressed in terms of spatial spread rather than position, enabling independent optimization of both position accuracy and reliability measurement without conflation.
3Adaptability or versatility
If the pixel group shape is used to calculate reliability, then the reliability becomes independent of subject shape, but the calculation complexity increases
Solution Approach 1:
The patent extracts the reliability information from the pixel group shape characteristics, specifically using the width of the peak region. By taking out only the essential shape parameter (peak width) rather than using the entire pixel group shape, the system achieves independence from subject shape while keeping calculation complexity manageable.
Data Source
AI summary
In order to facilitate generation of teacher data and to detect position coordinates with high reliability, there is provided a localization apparatus including: a deep learning model trained by using training image data in which position coordinates desired to be detected are specified and teacher image data in which a pixel group representing a shape independent of a subject of the training image data is arranged at a position relative to the position coordinates desired to be detected; a position coordinate calculation unit calculating position coordinates by using inference image data output from the deep learning model, and a reliability calculation unit calculating reliability by using global information of the pixel group of the inference image data output from the deep learning model.