Object Detection Training With Augmented Sensor Error States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in providing reliable object detection data due to faulty sensor readings, calibration issues, physical failures, and adverse weather conditions, which can compromise safety and operational reliability.
Innovation Solution
A system is developed that trains an object detection model using a data augmentation model to generate synthetic sensor data with error states, allowing the model to reliably detect objects even with faulty inputs, and incorporates a neural network for object detection and feature fusion to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor data is used directly for object detection, then the system operates in real-time, but the detection reliability deteriorates due to faulty sensor readings, calibration issues, and physical failures
Solution Approach 1:
The system performs data augmentation during the training phase to pre-teach the model about various error states and faulty sensor conditions. This preliminary action allows the model to learn robust feature extraction and fusion strategies before deployment, improving detection reliability without adding complexity to the real-time operation.
Solution Approach 2:
The system creates synthetic copies of sensor data with injected error states and fault conditions through data augmentation. These copied datasets simulate various failure scenarios (sensor malfunctions, calibration errors, adverse weather) allowing the model to train on diverse conditions without requiring additional physical sensors or complex hardware redundancy.
2Reliability
If data augmentation with error states is applied during training, then detection reliability under faulty conditions improves, but training data processing complexity increases
Solution Approach 1:
The system varies parameters of the training data by injecting different error states, noise levels, and fault conditions through data augmentation. This allows the model to learn invariant features across multiple parameter variations, improving robustness to faulty input data while maintaining a relatively simple training pipeline that builds upon existing object detection architectures.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A training device may include one or more processors configured to generate, using a data augmentation model, augmented sensor data for sensor data, the sensor data provided by a plurality of sensors, wherein the augmented sensor data comprise error states of one or more sensors of the plurality of sensors providing the sensor data, and to train an object detection model based on the augmented sensor data.