Managed data annotation · Philippines
Labeled data you can ship. Checked on every batch.
Send raw data and your guidelines. Every task goes through Attempt, Review and Final Review, and you get a quality report with each delivery.
- Attempt
- Review
- Final Review
Maria SantosPerson ordered twoQty Samsung monitorsProduct for the CebuLocation office on 12 MarchDate.
- 3 levels
- Attempt, Review and Final Review on every task
- Every batch
- seeded with hidden test items
- 500–1,000
- item paid pilot before production
- 1 business day
- to reply to every quote request
- Bounding boxes
- Polygons
- Keypoints
- Segmentation
- Named entities
- Sentiment
- Transcription
- Speaker diarization
- 3D cuboids
- Object tracking
- Intent tagging
- Label audits
01Services
One team for every data type your model learns from.
Labels for computer vision models, from simple tags to pixel-level masks.
OutputCOCO, YOLO, Pascal VOC or JSON
Full details on image annotation- Bounding boxes
- Polygons
- Keypoints
- Semantic and instance segmentation
- Image classification
02Quality
Every delivery comes with proof it meets the target.
We agree an accuracy target before work starts, then measure every batch against it.
Attempt, Review and Final Review. Nothing reaches you until a final reviewer has signed it off.
Items with known answers are mixed into every batch to measure accuracy.
We agree a target per project and report against it on every batch.
Work sent back at Review returns to the annotator once, with the reason, before going through Review again.
Accuracy on hidden test items
96.4%
Target 95%
- Items in batch
- 2,000
- Through Review
- 100%
- Through Final Review
- 100%
- Hidden test items
- 120
- Sent back at Review
- 4.8%
Ready for your acceptance · 7-day review window
03How it works
Five steps, on both sides of the work.
Every project starts with a scoped pilot. Production only begins once you've approved the results.
Request a QuoteTell us about your data, the labels you need, volume and deadline.
We agree guidelines, a label schema, a quote and an accuracy target.
A paid pilot of 500–1,000 items with 100% review, so you can check quality first.
Work runs in batches. Every task goes through Attempt, Review and Final Review, with hidden test items mixed in.
Each batch arrives with a quality report. Accept it or raise a dispute within 7 days.
04Industries
Guidelines shaped to the domain.
01
Autonomous vehicles
Cuboids, lane lines and tracked objects across camera and LiDAR data.
02
Retail and e-commerce
Product tagging, attribute extraction and shelf images for catalog and vision models.
03
Healthcare imaging
Segmentation and classification on medical images, under strict data handling.
04
Agriculture
Crop, weed and disease labels on field and drone imagery.
05
NLP and chatbots
Intent, entity and sentiment labels, plus review of model responses.
06
Research
Custom label schemas and careful guidelines for academic and lab datasets.
05Security
Annotators see the item in front of them. Nothing else.
- Client files stay in private storage. Annotators only see the items they are working on.
- No download access for annotators, and watermarking where you need it.
- Every annotator signs an NDA and contractor agreement before seeing data.
- Encryption in transit and at rest.
- Source files deleted after final delivery by default, or on your retention schedule.
For annotators
Remote work, paid per approved task.
Choose your hours. Get paid twice a month to GCash, Maya, PayPal or Wise. We invite people from the waitlist as client work comes in.
Join the annotator waitlist06Questions
What clients ask first.
Each project is quoted per item or per hour of data, based on the annotation type, volume and turnaround. We're taking on our first clients, so early projects get founding pilot pricing. Send a quote request and we will reply within 1 business day with a scope and price.
We reply to every quote request within 1 business day. Batch turnaround is agreed per project in the scope, based on volume, annotation type and your deadline. The pilot runs first, so you see real timing on your own data before production.
Yes. Every engagement starts with a paid pilot of 500–1,000 items, reviewed at 100%. You approve the pilot before production starts.
Every task passes three levels: an Attempt by an annotator, a Review by a reviewer, and a Final Review before it counts toward your batch.
Start with a pilot. Judge us on your own data.
We label 500–1,000 of your items with 100% review. You approve the results before production begins.
Founding pilot pricing
Annotask is taking on its first clients now. Early projects are quoted at founding pilot rates.
We reply within 1 business day. Want to send guidelines or a sample file? Use the full form.