Video Annotation

We need annotators to label short robotic manipulation videos using a strict timeline-based annotation tool. Each video must be fully covered from start to end with no gaps or overlaps. You will create two tracks: Subtasks (high-level steps) and Actions (fine-grained robot actions). For Actions, each annotation must represent one distinct, goal-directed action with accurate start/end timestamps (typically 0.5s–3s). Every action requires a structured 4-sentence description in English following an exact order: (1) observed state, (2) how the action relates to the overall task, (3) intended action, (4) expected outcome. You must write in first person, present tense, and use consistent robot arm naming (Arm A/B, Gripper A/B) with correct operator-perspective directionality (left/right/fore/aft). Some cases require point annotations to disambiguate identical objects/locations; when used, the text must reference the point using a specific keyword. You also must select an outcome rating for each annotation. This work is detail-heavy and quality-critical: small mistakes (wrong timing, missing/extra point, incorrect arm labeling, wrong sentence count/order, or gaps/overlaps) can fail the task. Applicants should be comfortable with repetitive high-precision work, strong written English, and careful following of rules and checklists. Required skills Excellent written English (clear, consistent, rule-following) High attention to detail (timestamps, no gaps/overlaps) Comfort working with a timeline-based UI Ability to follow strict formatting rules (exact sentence count/order) Patience and consistency (quality must remain stable across many videos) Nice to have Prior experience with data labeling / video annotation Familiarity with robotics or manipulation tasks (helpful but not required) You must do 4 tasks now, you should guarantee to pass all 4 tasks and then you will have 100 tasks

Project ID: 40302694