Case Study: Velodyne Lidar cuts training data selection time by 50% with Scale AI

Velodyne uses Scale Nucleus to find edge cases in 3D data and curate the most valuable data for annotation

Velodyne Lidar, a company that builds lidar sensors for safe navigation and autonomy, needed a better way to find useful training data within large sensor datasets. Its data team struggled to quickly identify rare edge cases for 3D sensor fusion and wanted an out-of-the-box solution for data selection and dataset management. Velodyne turned to Scale AI and Scale Nucleus to support this workflow.

Using Scale AI’s Nucleus platform, Velodyne queried 2D images tied to 3D point clouds with Object Autotag, model zoo tools, and natural-language search to quickly cluster and find similar scenes. This helped the team isolate rare examples in hours instead of days or weeks, speed up dataset creation, and start annotation jobs sooner. Scale AI also helped Velodyne reduce training data selection time by 50%.

Velodyne Lidar

Oliver Monson

Senior Manager, Data Operations