BlendTorch: A Real-Time, Adaptive Domain Randomization Library

TL;DR

BlendTorch is an adaptive library that generates infinite synthetic training data through real-time, randomized simulations, improving industrial object detection models more effectively than real datasets.

Contribution

We introduce BlendTorch, a novel adaptive domain randomization library that creates infinite synthetic data streams for real-time training in industrial computer vision tasks.

Findings

Models trained with BlendTorch outperform those trained on real data.

BlendTorch enables effective transfer of deep learning models to industrial environments.

Real-time data generation improves training efficiency and model robustness.

Abstract

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.