Autonomous Driving Road Scene Video Polygon Annotation Project
This project supported the training of L2-level autonomous driving perception algorithms. My core responsibility was polygon annotation for critical objects (vehicles, pedestrians, cyclists, traffic cones) in urban road scene videos. Over the 12-month project, I processed over 500 hours of video data, with an average of 30+ annotated objects per frame. I strictly followed ISO 9001-aligned annotation standards, using dual-review workflows and AI-assisted quality checks to ensure consistency. My annotation accuracy remained consistently above 99.5%, with zero critical annotation errors reported by the client.