Case Study: VirtuSense cuts false alarms and boosts model accuracy with Labelbox

VirtuSense cuts false alarms from 28% to 6% with Labelbox

VirtuSense, a company that provides an AI-powered fall-prevention system for healthcare facilities, faced a challenge in efficiently distributing unstructured data to its experts and tracking their progress. They struggled with open-source tools and a lack of support, which hindered their ability to produce the high-quality, diverse training data needed to reduce false alarms and increase model accuracy. This was critical for maintaining the trust of clinical staff who relied on their alerts.

By implementing Labelbox, VirtuSense built a self-retraining data engine that automated their model iteration loop. The Labelbox solution streamlined their workflow from data preparation to expert review and model retraining. As a result, VirtuSense increased its training data output by 5x, producing over one million evaluated assets. This led to a dramatic improvement in performance: false alarm rates dropped from 28% to 6%, and overall model accuracy increased by over 20%.

VirtuSense