Bounding box
A rectangle around each object. The most common type for object detection.
All services
Services · Data annotation
Your image archive is currently training nothing
Unlabeled data is functionally the same as no data. We annotate against your own taxonomy, with an AI pass first so people verify and correct instead of drawing from scratch — substantially faster, and priced accordingly.
Bounding box
Classification
Segmentation
Keypoints
Five annotation types
Pick by what your model needs to learn. More detail costs more effort, so unit prices differ — we itemise each type in the quote rather than collapsing it into one number.
Classification
One or more labels for the whole image. Cheapest and fastest, ideal for content sorting.
Segmentation
A polygon tracing the object outline. Needed when shape matters more than position.
Keypoints
Fixed landmark points on an object. Used for pose, joints and structure.
Vehicle damage
A specialised taxonomy: which part, what kind of damage, what severity, what action.
Three delivery models
Same taxonomy — the difference is who sits and reviews what the AI produced. This is the biggest lever on price, far bigger than negotiating unit rates.
Self-service
Cheapest · You keep full control
- You upload›
- AI pre-labels›
- Your team reviews
You use our annotation tooling with your own people. The AI does the rough pass; your team verifies and corrects. Right when the taxonomy is complex, or the data is sensitive and should not leave your team.
Crowdsourced
Balanced on price and speed
- You upload›
- AI pre-labels›
- Crowd reviews›
- Peer verification
TucTak's contributor network reviews the AI output, and uncertain cases go to multiple independent reviewers where only consensus passes. Right for volume work with a clear taxonomy.
Fully managed
Nothing for you to manage
- You upload›
- AI pre-labels›
- TucTak team works›
- TucTak signs off
A dedicated TucTak team does the work and owns the output quality. Right when the taxonomy needs domain knowledge (vehicle damage, documents) or you have nobody to run the process.
In and out
If your team uses a different format, send us one sample file — matching a schema usually takes half a day, and you do not have to write a converter.
- Images in JPG, PNG, WebP; video in MP4, MOV
- Upload through the dashboard or the API
- Returned as COCO, YOLO, Pascal VOC
- Or JSON against your own schema
- With a per-batch quality report
- Download through the dashboard or the API
How pricing works
Per image, not per hour. Three factors set the unit price — and all three are yours to choose, which is why quotes always itemise instead of collapsing into one number.
- Label type: classification costs a fraction of detailed segmentation
- Delivery model: self-service costs less than fully managed
- Volume: larger batches get better unit pricing
- No hard minimum — we take pilot batches of a few thousand images
Start with a small batch
Most clients begin with a few thousand images to judge quality themselves before committing volume. We encourage it — it is better for both sides.