Data Assets
Data assets are the complete set of uploaded data. In the asset list, each row is one episode, not one collection task. Every row can be traced to its collection task, device, format, scene, and curation status.
Where: Data Curation → Data Assets in the left navigation.
How Data Is Ingested
- When an upload completes, the platform runs integrity checks (gate G1): all channels are present, timestamps are continuous, and package counts match.
- Data that passes is ingested into the asset pool and assigned a data ID. Data that fails is set to Validation Failed and doesn't count toward task completion.
- Feedback archives (inference feedback and continuous operation data) go through the same ingestion process, with a different source tag.
- Structured outputs from pipeline processing are written to the asset pool automatically, tagged with the pipeline job ID.
Data IDs never change
A data ID is assigned at ingestion and stays fixed. All later curation, exports, and feedback reference the same ID.
Filter by Collection Task First
The first question people usually ask is "which collection run did this data come from?" The asset list therefore starts from the collection task: locate the task first, then narrow down step by step:
- Collection task (primary filter)
- Device (model / SN)
- Format (HDF5 / RAW)
- Scene
- Curation status
If a new task hasn't uploaded any data yet, the filtered list is empty and the page shows an empty-state message. This is expected.
List Columns
| Column | Description |
|---|---|
| Data ID | The unique identifier assigned at ingestion |
| Collection task | Collection task name and ID |
| Device | Model · name · SN |
| Format | HDF5 or RAW, matching the collection task's device type |
| Scene | The collection task's scene |
| Curation status | Ready to Curate / Awaiting Pipeline / Validation Failed / Exported |
Click Details to see the full metadata, including collector, collection time, duration, upload size, upload time, and storage path.
Four Asset States
| State | Meaning | Can be added to a curation project? |
|---|---|---|
| Ready to Curate | HDF5 data, or structured outputs from a pipeline | ✓ |
| Awaiting Pipeline | RAW data that hasn't been through a pipeline | ✗ |
| Validation Failed | Upload integrity checks did not pass | ✗ |
| Exported | Referenced by at least one dataset version | ✓ |
States change only as a result of actions and can't be edited manually.
Items awaiting a pipeline can't be selected
RAW items that haven't been through a pipeline are grayed out and can't be selected, not even with Select All. Run pipeline processing first, then curate the outputs once they are ingested.
One Asset, Many Projects
Different curation projects can reference the same read-only source asset, and items are deduplicated by data ID within a project. Labels and reviews apply only within the project where they were made and don't affect other projects.
Common Issues
Why isn't my data showing up in Data Assets? Check the task's successful upload count, package validation results, and retry history. Data is ingested only after it uploads successfully and passes validation; failures are not counted. Re-upload first, then investigate.
Why can't I select some items? Check whether they are in the Awaiting Pipeline or Validation Failed state.

