Multi-Vehicle Neural Network Training for Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in accurately recognizing objects across different images captured by multiple vehicles, leading to inconsistent object recognition rates and reduced safety in autonomous driving operations.
Innovation Solution
An electronic device and method that receive images and corresponding capturing information from multiple vehicles, determine if objects in these images are the same, and modify a neural network data recognition model using these images as training data to improve object recognition accuracy across vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple vehicles capture images independently with their own recognition systems, then each vehicle can operate autonomously, but object recognition accuracy becomes inconsistent across different vehicles
Solution Approach 1:
The patent merges image data from multiple vehicles to create a unified training dataset. By combining images captured by different vehicles under various conditions, the system creates a more diverse and comprehensive training set that improves recognition accuracy across all vehicles while ensuring consistency through shared learning.
Solution Approach 2:
The recognition model is designed to be universal across multiple vehicles. The same model trained on aggregated data from all vehicles is deployed to each vehicle, enabling the system to achieve both high accuracy and consistency across different vehicles through a single multi-functional model.
2Reliability
If a recognition model is trained on limited data from single vehicle, then training is computationally efficient, but recognition accuracy is insufficient
Solution Approach 1:
The patent combines image data from multiple vehicles to create a larger training dataset. This merging of data sources increases the quantity and diversity of training samples, enabling the model to learn from a broader range of scenarios and improve overall recognition accuracy.
Solution Approach 2:
The patent extends the training data dimensionality by incorporating spatial information from multiple vehicles. Instead of training on images from a single location and angle, the system utilizes images captured from different positions and perspectives across the vehicle fleet, adding spatial diversity to the training dataset.
3Reliability
If images from multiple vehicles are used for training, then recognition accuracy improves, but data processing complexity increases
Solution Approach 1:
The patent extracts and utilizes only the essential features from multiple images for training purposes. By focusing on key object characteristics and relevant image data while filtering out redundant information, the system reduces processing complexity while maintaining the benefits of multi-vehicle data aggregation.
Data Source
AI summary
An electronic device may include a processor configured to: obtain a plurality of images, the plurality of images including a first image captured by a first camera or image sensor, and a second image captured by a second camera or image sensor. The processor may further be configured to: determine whether a first object and a second object respectively included in the first image and the second image are a same object, and based on determining that the first object and the second object are the same object, train a neural network data recognition model to recognize the first object and the second object as the same object in both the first image and the second image by using the first image and the second image as training data of the neural network data recognition model.