Video Annotation Specialist & Labeling Assistant at Atlas Capture
Video
Action RecognitionServed as a Video Annotation Specialist and Labeling Assistant, producing high-granularity temporal labels for AI model training. Defined extremely precise action recognition temporal boundaries using highly granular timestamp increments down to 00.4 and 00.5 seconds. Enforced dense labeling protocols with strict segment duration and action-count constraints to preserve dataset consistency and model learnability. • Established and maintained temporal segmentation standards for action recognition tasks. • Applied dense labeling rules, including max 20-second segments and max two distinct actions per segment. • Performed QA and corrected model omissions by identifying skipped micro-actions. • Acted as a workflow subject matter expert, supporting improved evaluation standards.