Autonomous Vehicle Pedestrian Detection & Tracking
Performed high-precision labeling of 50,000+ video frames to train object detection models for self-driving cars. Tasks included drawing 2D bounding boxes for pedestrians and vehicles, as well as pixel-level semantic segmentation for road boundaries. Maintained a 98% quality accuracy rate through rigorous peer-review and consensus checks. Focused on edge-case scenarios like low-light conditions and occluded objects to improve model reliability.