I will robotics affordance labeling, keypoint annotation and embodied ai data

Generic bounding boxes are not enough for physical robot interaction. Embodied AI models require sub-pixel accuracy on grasp points, contact zones, and multi-point joint skeletons.

I specialize in high-precision robotics dataset labeling, converting raw video feeds and multi-camera spatial data into training-ready datasets for robotic end-effectors, hand-object interaction, and spatial movement models.

️ Specialized Services Offered

• Robotic Grasp & Contact Point Annotation: Precise tagging of grip zones, pinch points, and handle alignments for robotic grippers.
• Affordance Polygon Segmentation: Color-coded segmentation for interactable object parts (handle, lid, button, trigger, lever).
• 21-Point Hand Pose & Skeleton Tracking: Multi-frame keypoint annotation for human-robot interaction and teleoperation datasets.
• 3D Spatial Pose & Orientation Vector Tagging: Marking object rotation axes ($X, Y, Z$) and directional trajectory vectors across video frames.
• Multi-View Camera Synchronization: Labeling synchronized multi-camera feeds for spatial depth estimation.

Tools Expertise

• CVAT, Labelbox, Supervisely, Roboflow, Label Studio, Kitti, or client-proprietary web platforms.