Method for Operating an Artificial Intelligence System and System with an Artificial Intelligence System
Simple SummaryContent extracted from patent full text and abstract with AI.
This invention introduces a method and system for training and operating an artificial intelligence (AI) system using image data. The core idea is to train the AI with abstracted synthetic images and later apply it to real-world images that have been abstracted in a similar way. This enables the AI to better handle the differences between synthetic and real-world data, improving its performance when processing real images.
Use CasesContent extracted from patent full text and abstract with AI.
- Training AI models for image recognition or analysis where large datasets of labeled real images are unavailable or costly to obtain.
- Improving AI performance in applications such as automated quality inspection in manufacturing by using simulated images for training.
- Medical imaging analysis where synthetic data can be utilized for initial training and real patient images are later analyzed using the trained AI system.
- Autonomous vehicles, where simulated driving environments can be used to train perception systems before deploying them on real road images.
BenefitsContent extracted from patent full text and abstract with AI.
- Reduces reliance on large amounts of real-world image data for training, saving costs and time.
- Enhances the transferability and effectiveness of AI models from synthetic training environments to real-world scenarios.
- Improves AI system performance when handling the variability and complexity of real images.
- Enables faster development and deployment of AI solutions in fields where real data is limited or difficult to acquire.
Technical Classifications (CPCs)
Main Classifications
Physics & Measurement
Sub Classifications
Computing & Calculating
CPC Codes
Inventors & Applicants
Applicants
Friedrich Alexander Univ Erlangen Nuernberg Koerperschaft des Oeffentlichen Rechts
Schaeffler Technologies Ag
Patent Abstract
Die Erfindung betrifft ein Verfahren zum Betrieb eines künstliche Intelligenz-Systems (300),wobei das künstliche Intelligenz-System (300) - in einem Trainingsschritt - ausgehend von synthetisch erzeugten Bilddaten (105) trainiert wird, wobei die synthetisch erzeugten Bilddaten (105) mittels einer ersten Abstraktionsfunktion (106) abstrahiert werden, um synthetisch erzeugte, abstrahierte Bilddaten (107) zu erhalten und das künstliche Intelligenz-System basierend auf den synthetisch erzeugten, abstrahierten Bilddaten (107) trainiert wird, um ein trainiertes künstliche Intelligenz-System (300) zu erhalten;wobei - in einem dem Trainingsschritt (100) nachfolgenden Anwendungsschritt (200) - reale Bilddaten (201) bereitgestellt werden, die durch Aufnahme eines realen Bilds erzeugt sind, wobei die realen Bilddaten (201) mittels einer zweiten Abstraktionsfunktion (202) abstrahiert werden, um reale, abstrahierte Bilddaten (203) zu erhalten und die realen, abstrahierten Bilddaten (203) dem trainierten künstliche Intelligenz-System (300) zugeführt werden.
Key Information
Publication No.
DE102023116178A1
Family ID
93746302
Publication Date
2024-12-24
Application No.
DE102023116178A
Application Date
2023-06-21
Priority Date
2023-06-21
Granted
No
Possible Cooperation
For further information please contact the transfer office.