Bottom-Up Keypoint Association for Autonomous Vehicle Bounding Boxes

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing autonomous vehicle systems face challenges in accurately identifying keypoints and associating them with bounding boxes, particularly due to top-down keypoint detection methods that fail to provide accurate labels and are sensitive to variability in bounding box dimensions.

Innovation Solution

The system employs a bottom-up keypoint detection approach using a separate detection head within a machine learning model to identify keypoints such as vehicle corners, which are then associated with bounding boxes by identifying keypoints within defined bounding box extents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If top-down keypoint detection methods are used, then the system can process data efficiently, but the accuracy of keypoint labeling deteriorates and sensitivity to bounding box dimension variability increases

Engineering Contradiction:

Improvedata processing efficiencyVSAvoidkeypoint labeling accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent inverts the traditional top-down keypoint detection approach by implementing bottom-up keypoint detection. Instead of detecting keypoints from overall object detection results, the system first detects keypoints independently from sensor data and then associates them with bounding boxes. This inversion resolves the contradiction by achieving both high keypoint labeling accuracy and reduced sensitivity to bounding box dimension variability while maintaining processing efficiency.

Inventive Principle:

Principle #13The other way round (Inversion)

2Adaptability or versatility

If bounding box dimensions vary significantly, then the system can adapt to different object sizes, but the reliability of keypoint association deteriorates

Engineering Contradiction:

Improvebounding box size adaptabilityVSAvoidkeypoint association reliability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent changes the approach from relying on fixed bounding box dimensions to using learned parameters that adapt to varying object sizes. The bottom-up keypoint detection with association learning enables the system to reliably associate keypoints with bounding boxes across different scales by learning size-invariant features and association relationships, thereby maintaining reliability while preserving adaptability to different object sizes.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250078310A1Association of bottom-up keypoints based on bounding box extents

Publication Date: 2025.03.06 GM CRUISE HOLDINGS LLC

AI summary

Systems and techniques are provided for associating vehicle keypoints with a bounding box. An example method includes receiving, by a machine learning model configured to perform object detection, sensor data from one or more sensors of an autonomous vehicle; identifying, based on the sensor data, at least one vehicle that is located within an environment of the autonomous vehicle; generating a bounding box corresponding to the at least one vehicle, wherein the bounding box includes one or more bounding box extents that are based on a type of the at least one vehicle; identifying, based on the sensor data, a plurality of vehicle keypoints each corresponding to a vehicle feature; and associating, based on at least one bounding box extent of the one or more bounding box extents, one or more vehicle keypoints from the plurality of vehicle keypoints with the bounding box.