Ego Dextera
I worked on EgoDextera project, which involved annotating egocentric video data used to train computer vision models on hand-object interactions. My tasks included labeling actions, objects and interaction sequences at the clip and frame level, as well as reviewing and correcting existing annotations. The object covered thousands of short video clips from multiple participants. Quality was ensured through strict annotation guidelines, and regular quality checks, peer reviews, and feedback-based corrections to maintain high accuracy consistency.