Amazon SageMaker AI and Amazon Rekognition released a synthetic data augmentation pipeline for industrial safety AI, claiming it improves person detection accuracy. It is the company's first major update to its machine learning tools since 2023.
Amazon reported a 160 percent improvement in person detection mAP50 (mean Average Precision at an Intersection over Union threshold of 0.5) using synthetic data, measured on industrial safety AI benchmarks. That compares with previous results using real-world data.
The pipeline is built on Amazon SageMaker AI and targets industrial safety AI applications. Availability begins in 2024, initially for industries deploying autonomous equipment.
"We show how to use a synthetic data augmentation pipeline built on Amazon SageMaker AI and Amazon Rekognition to generate photo-realistic training images with automated labels," said the blog post. The approach uses real images with equipment but no people as substrates.
The announcement follows growing demand for safer industrial AI solutions. The source itself frames the significance as a step toward safer workplaces and more reliable AI models.
Amazon did not say how the synthetic data will be integrated with existing datasets, and the blog post raises the open question of how well the synthetic data generalizes to real-world scenarios. The company plans to continue refining the pipeline in the coming months.
Source: awsml
