AI Training Specialist - Video Annotation
Video
Bounding BoxAnnotated high-resolution video datasets to train computer vision models in detecting, tracking, and recognizing software interface elements and user interactions. Key Tasks Performed: - Applied frame-by-frame object tracking using 2d bounding boxes to accurately isolate software components, buttons, and dynamic UI changes. - Performed one-by-one elimination for complex multi-object frames to ensure zero omissions and maintain absolute structural precision across consecutive video frames. - Categorized and tagged localized UI attributes to help the ML model distinguish between different interactive states (e.g., active vs. inactive fields). Quality Measures: - Adhered to a strict Intersection over Union (IoU) threshold of >95% for bounding box tightness and alignment. - Delivered a 98%+ data accuracy rate through cross-verification and iterative self-review passes.