Case Study: CellarEye achieves 5x ML productivity and boosts average precision with Scale AI Nucleus

Delivering a Seamless Cellar Management Solution

CellarEye, Inc., a company that creates automated wine cellar management systems using computer vision, faced significant challenges in developing a reliable object detection model due to a complex environment and inconsistent data annotations, which hindered their model accuracy. To detect these data labeling errors and understand model failures, they turned to Scale AI and utilized its Nucleus platform.

By implementing Scale AI's Nucleus, the team could visually query their data, sort predictions by error metrics like IoU, and directly fix incorrect annotations. This solution enabled them to rapidly clean their data, which resulted in a dramatic improvement in model performance; one specific class saw its average precision jump from 76% to over 87%. Furthermore, Scale AI's tools provided a 5x increase in ML productivity efficiency and delivered 5x savings on annotation review costs for CellarEye.

CellarEye, Inc.

Mehdi Mohseni

Co-Founder and CEO