Add action labels, temporal segments, trajectories, contact states, and customer-defined metadata against an agreed training or evaluation schema.
Skeletal animation records how a body moves. It does not necessarily identify the action, mark where one phase ends and another begins, describe ground or object contact, or use the field names a model expects.
Uthana scopes labeling and enrichment around an explicit schema. The source can be studio-captured motion, motion estimated from video, or customer-provided skeletal data, provided the required fields can be observed or derived reliably from that source.
Not every label is valid for every source. A video-derived sequence may carry different uncertainty than marker-based capture. If a requested field cannot be produced reliably, that limitation is identified during sampling rather than hidden inside a low-confidence delivery.
Programs can be scoped for:
The schema, allowed values, granularity, sampling, and exception behavior are agreed before scaled production.
Each engagement begins with source data, a target schema, and written acceptance criteria. Uthana produces a representative sample so the customer can verify field definitions, sequence boundaries, value ranges, file integrity, and pipeline compatibility before scaling.
Automated checks can validate structural requirements and allowed ranges. Semantic labels and ambiguous motion may require human review. The final QA process, review rate, exception handling, and acceptance threshold are documented for the engagement.
This sample validates the delivery against its specification. It does not by itself prove that the data will improve a downstream model.
Share the motion source, skeleton, required fields, examples, and acceptance checks. If the schema is still forming, begin with the smallest sample that can expose ambiguity and technical constraints.
Start generating motion today, or work with us on a custom dataset, embodiment system, or simulation program.
