Pegasus 1.6 Turns POV Video Into Robot Data

Pegasus 1.6 makes video labeling a more credible entry point for robot-data infrastructure, though generated annotations still need validation before controlling real systems.

• –Supports first-person footage that general-purpose video models often mishandle
• –Produces structured outputs for action labeling, quality checks, and trajectory prediction
• –Native image analysis lets teams reuse one API across video and still-image workflows
• –Improved entity recognition and persistent metadata should reduce brittle downstream processing
• –The real test will be annotation accuracy across diverse robots, operators, and edge cases