EgoWild-X is a large-scale dataset of egocentric video sequences capturing dexterous, in-the-wild human interactions with ground-truth hand motion using the Manus MetaGloves Pro. Using head-mounted wide-lens GoPro cameras (270° FOV) synchronized with glove-based joint tracking, the dataset provides high-resolution, multimodal recordings ideal for training embodied agents with fine motor control. This is similar to our EgoWild dataset, with the addition of motion capture.

Unlike lab-constrained datasets, EgoWild-X emphasizes real-world household and assembly tasks performed in natural settings. Recordings include:

• Cooking (chopping, stirring, opening jars, pouring)

• Cleaning & Laundry (wiping, folding, ironing, vacuuming, picking up clutter)

• Assembly & Repair (building furniture, screwing parts, tightening bolts)

• Daily Routines (typing, writing, opening doors, using laptops)

• Utility (unpacking, carrying bags, moving furniture)

Objects range from deformables (clothes, sponges, food) to rigid/articulated items (drawers, appliances, tools), with recordings spanning kitchens, garages, offices, and outdoor environments—complete with clutter, occlusion, and social presence.

EgoWild-X supports research in:

• Learning dexterous manipulation policies from egocentric multimodal data

• Understanding object affordances in realistic, task-driven contexts

• Bridging vision-based hand pose estimation with ground-truth kinematics

• Advancing generalist robot hands for everyday environments

The task-rich, untrimmed nature of EgoWild-X makes it uniquely suited for embodied AI systems that must learn from the same kinds of activities humans perform daily.