Adaptive-Resolution Semantic Occupancy Maps for Lower Compute Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
3D semantic occupancy prediction is challenging due to high computational complexity and memory requirements associated with processing large-scale 3D data.
Innovation Solution
An adaptive resolution generator that uses a dual-stage approach involving a lower-resolution module and a higher-resolution module to generate hybrid occupancy maps, selectively focusing computational resources on relevant regions for enhanced detail and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution is used for the entire 3D occupancy map, then measurement precision is improved, but computational complexity increases
Solution Approach 1:
The patent applies local quality by generating high-resolution occupancy maps only for specific regions of interest (such as areas containing objects or important features) while using lower resolution for the remaining background regions. This allows the system to maintain high measurement precision where needed while reducing overall computational complexity and memory requirements.
Solution Approach 2:
The patent segments the 3D occupancy map into multiple resolution levels or regions, processing different portions at different resolutions. This segmentation strategy divides the computationally intensive task into manageable parts, allowing high precision processing only for critical areas while using coarser processing for less important regions.
2Measurement precision
If high resolution is used for the entire 3D occupancy map, then measurement precision is improved, but memory requirements increase
Solution Approach 1:
The patent implements local quality by allocating high memory resources only to specific regions requiring detailed representation while using minimal memory for background areas. This selective memory allocation maintains high measurement precision for important regions while significantly reducing overall memory requirements compared to a uniformly high-resolution approach.
Solution Approach 2:
The patent segments the occupancy map data structure into multiple resolution levels, storing detailed high-resolution data only for regions of interest while using compressed or downsampled representations for other areas. This segmentation reduces the total quantity of data that must be stored in memory while preserving necessary precision.
3Productivity
If low resolution is used for the entire 3D occupancy map, then computational complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent applies local quality by using low resolution processing for background or less important regions to maintain high processing speed, while simultaneously applying high resolution processing to regions containing objects or features of interest to maintain adequate measurement precision. This hybrid approach balances productivity and precision based on spatial location.
Solution Approach 2:
The patent segments the processing task into multiple resolution passes or regions, performing rapid low-resolution processing for most areas while applying detailed high-resolution processing only to specific segments containing important features. This segmentation enables the system to achieve overall high productivity while maintaining necessary precision where required.
Data Source
AI summary
A computer-implemented method and system include a first machine learning system, which generates feature maps using a set of digital images. A second machine learning system uses the feature maps to generate object boundary data of a set of objects, which are displayed in the set of digital images. Three-dimensional (3D) feature volume data are generated using the feature maps. A coarse occupancy map is generated using the 3D feature volume data. The coarse occupancy map has a first resolution. The coarse occupancy map includes an environment and the set of objects. Surface data is generated using the object boundary data and the 3D feature volume data. The surface data has a second resolution. A hybrid occupancy map is generated by combining the coarse occupancy map and the surface data. The hybrid occupancy map displays the environment with the first resolution and the set of objects with the second resolution.