Artificial intelligence (AI) and machine learning are reshaping the aerospace and government sectors and simulation and training community. Yet for computer vision applications, data – the key ingredient for machine learning training – present significant challenges.
Real-world data is expensive to acquire and often prohibited to collect for security reasons. Additionally, annotating data is both expensive and prone to human error. Real-world data can also only prepare for what has already happened, not future scenarios. In an industry like defense where the emphasis is on preventative measures, that doesn’t cut it.
In this talk delivered at the TEAR 3M 2021 event, Joe Mercado, Ph.D., a senior product manager, AI at Unity, covers:
- The challenges of training deep learning models
- How synthetic data overcomes data collection and annotation challenges
- How computer vision tools from Unity generate labeled, high-quality data for machine learning training cost-effectively
- An aviation case study showcasing how augmented reality (AR) and synthetic data were used to improve maintenance procedures
Learn more