Learning goal
After watching, you can take an approved masked output from MaskingOps and package it for long-term archival with FAIR-compliant metadata. You know what gets archived openly, what stays restricted, and what paperwork the repository will require.
Transcript
Cold open
End of the pipeline. The masked video has been benchmarked, reviewed, and approved. What stops it from sitting on a shared drive forever is this step — packaging it so someone else can actually find, access, and reuse it.
What FAIR actually demands
FAIR is four letters. Each one is a question your archive needs to answer with evidence — not adjectives.
Findable. Does it have a persistent identifier — a DOI from DANS, DataverseNL, or Zenodo? If your only identifier is a folder path on a server, that's not findable.
Accessible. What can people get, and how? Three common settings — fully open, controlled access for authorised researchers, or metadata-only with the data restricted.
Interoperable. Is the metadata written in a standard schema — Dublin Core, DDI — so other systems can read it, not just yours?
Reusable. Clear licence. Documented provenance. The next researcher can take this and use it without writing to you to ask what the masking method was.
What goes in the deposit package
Four things in the package, plus the videos themselves.
1. The masked videos. From MaskingOps, signed off. Original videos do not go in this package — they stay restricted in your secure institutional storage.
2. A metadata file (Dublin Core / DDI). Use the SYNAPSIS metadata schema template — it covers what behavioural-AV datasets need: title, creators, masking method, software version, body regions masked.
3. A masking report. The processing parameters, the QA log from MaskingOps, the sampling protocol. This is what makes the masking reproducible. Template available in the SYNAPSIS downloads folder.
4. Data-sharing decision-tree result. The output showing why this dataset is being shared at this access level — open, controlled, or metadata-only. The tree itself is in the same downloads folder.
Choosing the access level
Three access tiers. Pick the one that matches your data sensitivity and consent.
Open. Anyone with the DOI can download. Reserved for fully-masked data with broad participant consent and low residual identification risk. Think: skeleton-only outputs of public-figure recordings.
Controlled access. Researchers apply, sign a data-use agreement, get authorised. The DANS Data Station SSH supports this directly. This is the realistic default for most behavioural AV data.
Metadata-only. The metadata record is public — title, abstract, methods. The data itself stays restricted to the original team. Still FAIR-findable. Still cite-able. Still useful for meta-analysis at the description level.
The DANS deposit, step by step
For Dutch SSH research, DANS Data Station SSH is the default destination. Other options — DataverseNL, Zenodo — work too; the workflow is similar.
Step 1. Reserve a DOI before your paper goes out. Include it in the manuscript so the reference resolves the day the paper is published.
Step 2. Fill in the metadata form using the SYNAPSIS schema template as a checklist. The repository's fields map to Dublin Core / DDI; the SYNAPSIS template tells you what to put in each.
Step 3. Upload the deposit package — masked videos, masking report, decision-tree result. Set the access tier you chose.
Step 4. Submit. The repository reviews the deposit — they check the metadata is complete, not the science. Typically a few days to two weeks.
Recap
Archiving is the four FAIR letters answered with evidence. Findable — DOI. Accessible — pick a tier. Interoperable — Dublin Core or DDI. Reusable — licence, provenance, masking report.
That's the full SYNAPSIS pipeline. MaskAnyone, MaskBench, MaskingOps, Archiving. From sensitive recording to citeable FAIR dataset. For the mechanics — installation, hands-on first job, quality assessment in detail, troubleshooting — see the deep-dive companion tutorials.