Multi-Resolution Top-Down Object Classification With Split ML Pipelines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for processing sensor data in autonomous vehicles face challenges in resource-constrained environments due to the need to handle substantial input data, which can lead to processing restraints and inefficiencies.
Innovation Solution
The approach involves separating processing pipelines for different ranges of multi-channel image data, with a first portion of the ML model processing spatial features and a second portion processing non-spatial features in parallel, and then aggregating the results to determine classification probabilities, allowing for efficient data representation and reduced processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If substantial input data is processed to determine object classifications, then measurement precision and reliability are improved, but processing time increases and resource requirements increase
Solution Approach 1:
The patent segments the processing pipeline into multiple portions (e.g., first portion for initial processing, second portion for refined processing) that operate in parallel. Different resolutions of the same input data are processed simultaneously through different pipeline portions, allowing the system to maintain high classification accuracy while reducing overall processing time by avoiding sequential processing bottlenecks.
2Measurement precision
If high-resolution data is processed for distant objects, then measurement precision is improved, but use of energy and computational resources increases
Solution Approach 1:
The patent applies local quality by processing different portions of the input data at different resolutions based on their specific requirements. Distant objects that require higher precision for safe detection are processed at higher resolutions through specific pipeline portions, while closer or less critical objects are processed at lower resolutions. This selective approach maintains necessary detection accuracy for safety-critical distant objects while reducing overall computational resource consumption.
Data Source
AI summary
Techniques for determining a classification probability of an object in an environment are discussed herein. Techniques may include analyzing sensor data associated with an environment from a perspective, such as a top-down perspective, using multi-channel data. From this perspective, techniques may determine channels of multi-channel input data and additional feature data. Channels corresponding to spatial features may be included in the multi-channel input data and data corresponding to non-spatial features may be included in the additional feature data. The multi-channel input data may be input to a first portion of a machine-learned (ML) model, and the additional feature data may be concatenated with intermediate output data from the first portion of the ML model, and input into a second portion of the ML model for subsequent processing and to determine the classification probabilities. Additionally, techniques may be performed on a multi-resolution voxel space representing the environment.