Dynamic Knowledge Base for Neural Network Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in identifying objects in real-time due to limited computing systems and the need for frequent manual retraining with new data, making it difficult to maintain accurate object detection at higher speeds.
Innovation Solution
A method and device that segment scene images into pixels, detect objects, find similar images in an online database, identify objects using a knowledge base engine, update the knowledge base, and train a neural network for improved object detection, enabling dynamic learning and accurate navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional one-time training with limited data is used, then device complexity is reduced, but object detection accuracy deteriorates due to inability to adapt to new objects
Solution Approach 1:
The system transitions from static one-time training to dynamic continuous learning. The control system automatically retrieves new data from online databases, updates the neural network, and re-trains itself in real-time during vehicle operation, enabling adaptation to new objects without manual intervention
Solution Approach 2:
The control system performs self-updating and self-training by automatically collecting new object data, updating its knowledge base, and re-training the neural network without external manual training. The system serves itself by continuously improving its own detection capabilities
2Measurement precision
If large data set collection and processing is performed, then object detection accuracy is improved, but productivity deteriorates due to limited computing system speed
Solution Approach 1:
The system pre-collects and stores large datasets in online databases before they are needed for training. When training is required, the pre-prepared data is quickly retrieved and processed, eliminating the time-consuming data collection and processing steps during actual operation
Solution Approach 2:
Online databases serve as an intermediary between data sources and the control system. The databases store, organize, and manage large datasets, allowing the control system to access processed training data without performing heavy data collection and preprocessing tasks itself
3Measurement precision
If manual training is performed frequently, then object detection accuracy is improved, but loss of time increases due to repeated training cycles
Solution Approach 1:
The control system automatically performs data collection, processing, and model re-training without human intervention. This eliminates the time required for manual training operations while maintaining continuous improvement of detection accuracy through automatic adaptation to new objects
Solution Approach 2:
The system implements continuous learning by constantly retrieving new data, updating the neural network, and re-training during vehicle operation. This continuous process eliminates gaps between training cycles, ensuring the system remains up-to-date without periodic manual intervention
Data Source
AI summary
A method and device for detecting objects from scene images by using dynamic knowledge base is disclosed. The method includes segmenting a scene image captured by at least one camera into a plurality of image segments. Each of the plurality of image segments include a plurality of pixels. The method further includes detecting at least one object in each of the plurality of image segments. The method includes finding a plurality of similar images from an online database based on the at least one object. The method further includes identifying using a knowledge base engine, at least one similar object in the plurality of similar images. The method includes updating the knowledge base engine based on the at least one similar object identified from the plurality of similar images. The method further includes training a neural network to detect objects from scene images, based on the updated knowledge base engine.