Annotation Tasks
Coming Soon!
Annotation Tasks are an accessible way for users to continuously contribute to the EgoPlay ecosystem while helping improve Orn’s understanding of human activity. Unlike video submissions, which generate new data, annotation tasks focus on refining and labeling existing videos - turning raw footage into structured, machine-readable datasets for robotics training.
Purpose
Every day, thousands of egocentric videos are submitted to EgoPlay. Orn processes these videos automatically, but human input remains essential for edge cases, context validation, and nuanced labeling that AI alone can’t yet handle.
Annotation Tasks provide that human feedback loop - ensuring that every video in the ecosystem is properly categorized, labeled, and ready for use in robotics model training. In doing so, contributors help make Orn smarter, faster, and more contextually aware over time.
Task Flow
Assignment – Users are presented with short clips or image frames from videos within the EgoPlay system.
Annotation – The user either answers simple multiple-choice questions or applies predefined labels (e.g., “Cutting,” “Folding,” “Pouring,” “Cleaning”).
Submission – Once labeling is complete, the task is automatically submitted to Orn for verification.
Reward – Upon successful verification, users earn Vader Points (VP).
Annotation Tasks are intentionally designed to be fast, simple, and repeatable, enabling anyone to meaningfully participate in the AI training process without requiring technical expertise.
Energy & VP Rewards
Energy Cost: Each Annotation Task consumes 1 Energy ⚡️.
Unlimited Tasks: Users can complete an unlimited number of annotation tasks per day as long as they have Energy available.
VP Rewards: VP Rewards for successfully completed annotations tasks will be lower than for successfully completed video tasks.
Why Annotation Matters
Annotation Tasks form the connective tissue between user participation and AI learning. Each label improves Orn’s models, helping refine its ability to detect, segment, and understand complex human actions from a first-person view.
By crowdsourcing this process through an open, Energy-based system, EgoPlay makes large-scale, high-quality labeling possible - faster, cheaper, and more diverse than traditional data pipelines.
In short, annotation turns every user into an active trainer of embodied AI, accelerating the progress of robotics development worldwide - one labeled clip at a time.
Last updated