Data curation
Deepen AI's trained teams go through your raw logs and recordings to find gaps, duplicates and edge cases, so you train on the data worth using.
Rate in your plan before you pay A plan within 2 working days No sales call
What a curation decision looks like
Each clip or log comes back kept, dropped or flagged, with the reason and its tags.
- Clip
- Drive 3301, 40 seconds of highway on a clear day
- Your filter marked
New scene.- Curated as
- Dropped: near duplicateMatches 212 clips already kept from the same road and conditions.
- Reason
- Your rules cap clear-day highway clips at 200.
- Clip
- Drive 4128, 12 seconds at a city junction
- Your filter marked
Routine.- Curated as
- Kept: edge caseA cyclist rides the wrong way through the junction at 0:07.
- Reason
- Tagged “wrong-way road user”, a scenario with only 9 clips so far.
- Collection
- Warehouse pick logs, week 41
- Your filter marked
Complete.- Curated as
- Gap foundNo picks of transparent items: 0 of 1,200 runs.
- Reason
- Your coverage list asks for at least 50. The gap is added to the next collection spec.
- Clip
- Drive 5002, 20 seconds through a construction zone
- Your filter marked
- Edge case.
- Curated as
- Sent to youThe zone is closed and empty. Does a closed construction zone count as an edge case?
- Reason
- Your answer is added to the curation rules.
How we start on your work
No sales call. A plan in your inbox, then a short paid pilot on your real logs.
What we need from you
- Your logs or recordings, and where they're stored
- Your coverage list: what you need more of, and less of
- Your tags or scenario categories, if you have them
- 5 minutesTell us the task
What data you have, how much, and what you're short of.
- Within 2 working daysGet a plan and a price
People, hours, the rate, start date and the total price.
- 2 weeksRun the pilot
We train the team on your rules, curate real logs and send two weekly reports. Then you decide.
Who curates, and how decisions stay consistent
Trained reviewers, working in English. Each one learns your coverage list and tags and curates a set of known clips before live work.
Every reviewer's accuracy is measured
Items you've already answered are hidden in the real work, so we know how accurate each person is.
Nobody guesses on unclear cases
The team lead decides, or asks you. Your answer joins the guidelines.
You see the results every week
A weekly report shows work done, QA results and what's improving.
Standard curation covers camera, LiDAR and telemetry recordings and robot logs, in your viewer or Deepen AI's Curate. Annotating the kept data is quoted as an enterprise program.
How much does data curation cost?
Data curation is quoted in your plan, because the rate depends on your data and the tools it's viewed in. You pay for the hours people work, with a team lead, QA checks and a weekly report included.
Set before you pay. Your plan shows the rate, team size, hours and total price.
Starts with a two-week pilot on your real logs, paid upfront, with two weekly reports.
Your data stays protected. We work under NDA with the access you set. Deepen AI holds SOC 2 Type II, ISO 27001, ISO 9001 and TISAX.
Questions
We already filter automatically. Why add people?
Filters find what they're built to find. People notice the unusual: the wrong-way cyclist, the odd reflection, the scene no rule describes. Many teams filter first and have people curate what's left.
What happens to the data you drop?
It stays where it is. Curation marks each item kept or dropped with a reason, and you decide what to remove.
Can you show what we have, by scenario?
Yes. The weekly report counts your data by scenario and condition, and lists the gaps against your coverage list.
Do you work in Deepen AI's Curate?
Yes, or in the viewer your team already uses. Your plan says which.
What format are decisions delivered in?
Your schema, such as a CSV or JSON list with the item ID, keep or drop, the reason and the tags.
Can gaps turn into new data?
Yes. Gaps can go straight into a collection spec, so the next sessions record what's missing.