Case Study: Protex AI achieves 5x faster model iteration with Voxel51's FiftyOne

Protex AI uses FiftyOne to intelligently balance computer vision datasets and improve model accuracy

Protex AI, a provider of AI-powered workplace safety solutions, sought to enhance and scale its computer vision pipeline for analyzing footage from over 100 industrial sites. Their challenge involved managing fragmented tools for data validation, curation, and model analysis. To overcome this, they adopted the FiftyOne platform to create a more streamlined and collaborative ML operations hub.

By implementing FiftyOne, Protex AI consolidated its workflow into a single, flexible tool. The solution utilized FiftyOne's SDK and plugin framework to build roughly ten customized workflows for tasks like data filtering and running inference. This integration resulted in a 5x faster model iteration and enabled the processing of data from over 1000 cameras, ultimately helping their clients achieve over 80% reductions in workplace incidents.

Protex AI

Patrick Rowsome

Lead Computer Vision Engineer