Bounding Box Selection via Machine Learning Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous systems face challenges in accurately representing and tracking three-dimensional objects in environments due to the generation of multiple bounding boxes with variations in size, shape, orientation, and confidence levels, which complicates object tracking and navigation.
Innovation Solution
A machine learning model is trained to select or generate a bounding box that most accurately represents an object by processing sensor data annotated with ground truths, considering characteristics such as velocity, proximity to road features, and confidence levels, allowing for the selection of the most appropriate bounding box type or generation of a new bounding box based on specific situations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple three-dimensional bounding boxes are generated for an object, then the representation of the object becomes more comprehensive, but the complexity of object tracking and processing increases
Solution Approach 1:
A machine learning model is introduced as an intermediary component that receives multiple candidate bounding boxes as input and outputs a single selected bounding box. This intermediary resolves the contradiction by comprehensively evaluating multiple representations while simplifying the output to a single box for tracking, thus maintaining reliability without increasing downstream complexity
Solution Approach 2:
The system changes the parameter state of bounding boxes by evaluating multiple candidate boxes with different characteristics (position, size, orientation, confidence levels) and transforming this multi-parameter input into a single optimized bounding box output, resolving the complexity issue while preserving comprehensive object representation
2Device complexity
If a single three-dimensional bounding box is used for object tracking, then the processing complexity is reduced, but the accuracy of object representation may be compromised
Solution Approach 1:
The system performs preliminary action by generating and evaluating multiple candidate bounding boxes before final selection. This pre-processing step ensures that the single bounding box used for tracking is the most accurate representation, thus maintaining measurement precision while keeping downstream processing simple
Solution Approach 2:
The machine learning model performs self-service by autonomously evaluating multiple candidate bounding boxes and selecting the optimal one without requiring external intervention. This self-contained process ensures accurate object representation while maintaining simple processing architecture
Data Source
AI summary
Techniques to train a model with machine learning and use the trained model to select a bounding box that represents an object are described. For example, a system may implement various techniques to generate multiple bounding boxes for an object in an environment. Each bounding box may be slightly different based on the technique and data used. To select a bounding box that most closely represents an object (or is best used for tracking the object), a model may be trained. The model may be trained by processing sensor data that has been annotated with bounding boxes that represent ground truth bounding boxes. The model may be implemented to select a most appropriate bounding box for a situation (e.g., a given velocity, acceleration, distance, location, etc.). The selected bounding box may be used to track an object, generate a trajectory, or otherwise control a vehicle.