Grid-Generated CNN Parameter Mapping for Off-Center Road Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Convolutional Neural Networks (CNN) used in autonomous driving struggle when input images deviate from the typical arrangement of training images, leading to inefficient processing due to optimized parameters being specific to roads in the center, which can result in improper handling of roads in different positions in test images.
Innovation Solution
A method using a grid generator to divide test images into subsections based on dynamic templates, adjusting boundaries to include pairs for testing, and selecting parameters for neural network operations by comparing classes and areas of objects and non-objects within these subsections, allowing for optimized parameter determination and improved processing even when image arrangements differ from training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CNN parameters are optimized for training images with typical arrangement (roads in center), then processing efficiency is improved for those specific arrangements, but processing accuracy deteriorates when test images have different arrangements (roads not in center)
Solution Approach 1:
The test image is divided into multiple subsections (e.g., center, left, right, upper, lower regions) using a grid generator. Each subsection is independently processed by selecting CNN parameters that match its specific content type and arrangement, rather than applying uniform parameters to the entire image. This segmentation allows the system to handle diverse road arrangements effectively.
Solution Approach 2:
Different CNN parameters are selected for different subsections based on their local characteristics. The system determines parameters for each subsection by comparing the content (road, sky, sidewalk, etc.) and arrangement in that specific region with training data patterns, enabling locally optimized processing that adapts to varying image arrangements across different regions.
2Device complexity
If uniform CNN parameters are used for all regions of the image, then device complexity is reduced, but adaptability to different image arrangements deteriorates
Solution Approach 1:
The system pre-establishes multiple sets of CNN parameters corresponding to different image arrangements and content types (road in center, road on left, road on right, etc.). During processing, the grid generator and parameter selector quickly match the current subsection's characteristics with the pre-prepared parameter sets, avoiding complex real-time optimization while maintaining high adaptability.
Solution Approach 2:
The system dynamically changes CNN parameters based on the detected content and arrangement in each subsection. By comparing the class information and spatial distribution of objects in each subsection with training data patterns, the system selects the most appropriate parameter set for that specific region, enabling flexible adaptation without increasing overall system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of neural network operations by using a grid generator is provided for converting modes according to classes of areas to satisfy level 4 of autonomous vehicles. The method includes steps of: (a) a computing device instructing a pair detector to acquire information on locations and classes of pairs for testing by detecting the pairs for testing; (b) the computing device instructing the grid generator to generate section information by referring to the information on the locations of the pairs for testing; (c) the computing device instructing a neural network to determine parameters for testing by referring to parameters for training which have been learned by using information on pairs for training; and (d) the computing device instructing the neural network to apply the neural network operations to a test image by using each of the parameters for testing to thereby generate one or more neural network outputs.