The problem
What goes wrong without it.
Where training data usually breaks — and what this service is designed to prevent.
- Bounding boxes are fast but include background; for irregular shapes the model learns the wrong pixels.
- Keypoints and skeletons are one of the most skilled annotation disciplines — occluded joints still need a decision.
- Common errors — missing boxes, loose boxes, wrong classes or attributes — transfer straight into model behaviour.
Scope
What is included.
Bounding boxes
The most widely used technique: object location and class, efficient at scale.
Polygons
A precise envelope around irregular objects — aerial imagery, AR/VR, defects.
Semantic segmentation
Pixel-accurate masks for scene understanding, per class or full frame.
Keypoints & skeletons
Joints, facial landmarks and hand points for pose, emotion and gesture recognition.
Polylines
Lanes, edges, cables and other linear structures.
Classification
Image-level classes and attributes, including multi-label tagging.
What you get
- Annotations in COCO JSON, YOLO, Pascal VOC, CSV or CVAT XML
- Taxonomy and edge-case guidelines (occlusion, truncation, minimum size)
- Inter-annotator agreement and QA results
- A recommendation on annotation type when the task allows several
Starting prices
- 2D bounding boxCamera€0.04 / box
- PolygonCamera€0.05 / polygon
- KeypointCamera€0.02 / keypoint
- Full semantic segmentationCamera€3.00 / frame
Starting prices in EUR, net of VAT, with two QA passes included. Your final unit price depends on object density, number of classes and attributes — we confirm it after a sample.
All prices, pilot and NDAFAQ
2D Annotation questions.
Bounding boxes or polygons — which should I use?
Boxes are enough for counting and most detection tasks. Choose polygons or segmentation masks when the object shape matters — irregular objects, defects, land parcels — because boxes include background.
How do you handle occluded keypoints?
We agree a rule before production — for example label with a visibility flag or skip — and write it into the guidelines with examples so every annotator applies it the same way.
What does 2D annotation cost?
Starting prices are €0.04 per bounding box, €0.05 per polygon, €0.02 per keypoint and €3.00 per fully segmented frame, with two QA passes included. The final price is confirmed after a sample.
Further reading
From the BoBox blog.
Keypoint Annotation When Boxes Are Not Sufficient!
When bounding boxes aren't enough, keypoint annotation enables pose estimation, gesture recognition, and emotion detection. Here's everything you need to know.
Why Data Annotation is Our Mission Critical?
Data annotation is a critical component of any AI project, particularly in computer vision, where precise labeling of images directly impacts model performance.
The Importance of Details in Data Annotation
Should a half-visible person be labeled? The answer depends on what the model is for — and small decisions like this make or break computer vision projects. Here's how to get annotation details right.
2D Annotation, done right the first time.
Tell us about your data, model and timeline. A senior team member will reply within 72 hours.
