BlendTorch: A Real-Time, Adaptive Domain Randomization Library
This addresses the data scarcity issue for industrial applications of computer vision, offering a practical solution to enable the transfer of state-of-the-art methods.
The paper tackles the problem of insufficient real-world training data for deep learning in industrial computer vision by introducing BlendTorch, an adaptive Domain Randomization library that generates infinite synthetic data streams, resulting in models that outperform those trained on real or photo-realistic datasets in industrial object detection tasks.
Solving complex computer vision tasks by deep learning techniques relies on large amounts of (supervised) image data, typically unavailable in industrial environments. The lack of training data starts to impede the successful transfer of state-of-the-art methods in computer vision to industrial applications. We introduce BlendTorch, an adaptive Domain Randomization (DR) library, to help creating infinite streams of synthetic training data. BlendTorch generates data by massively randomizing low-fidelity simulations and takes care of distributing artificial training data for model learning in real-time. We show that models trained with BlendTorch repeatedly perform better in an industrial object detection task than those trained on real or photo-realistic datasets.