Data labeling
Image annotation services for computer vision models
Bounding boxes, polygons, pixel-accurate masks and keypoints, drawn by trained annotators in Nairobi and checked by a second reviewer before delivery. Your classes, your edge-case rules, exported in the format your pipeline expects.
- Pilot phase
- Two or more review layers
- You own every label
Labels tight enough to train on.
Computer vision models learn the exact boundary you draw. A box that is four pixels loose, a polygon that skips the side mirror, or a class swapped on a lookalike object all become patterns your model repeats. Our image annotation teams are trained on your guidelines and tested on a gold set before they label production data.
Every image passes at least two review layers: a peer check on each task and a QA audit on a sample of every batch. Ambiguous cases are escalated to your team instead of guessed, and the answer is added to the guideline so the same question never comes up twice.
Used for
- Object detection for retail shelves, logistics and security
- Medical imaging segmentation under strict access controls
- Crop, weed and disease detection for agritech
- Satellite and drone imagery for mapping
- Defect detection on manufacturing lines
Industries: Retail & e-commerce · Healthcare & medical imaging · Agriculture & agritech · Manufacturing & robotics
What our image annotation covers
Combine any of these in one workflow and one guideline.
Bounding boxes
2D boxes with classes, attributes, occlusion and truncation flags for object detection.
Polygons
Instance outlines that follow the real shape of objects, for irregular items and crowded scenes.
Semantic & instance segmentation
Pixel-level masks for every class, with clean edges and consistent handling of overlaps.
Keypoints & pose
Body, hand, face and object landmarks with visibility flags for hidden points.
Image classification
Single and multi-label tags, quality filtering and dataset curation.
Polylines & lanes
Lane markings, road edges, cables and other thin structures.
Formats and tools
Delivery formats
- COCO JSON
- YOLO
- Pascal VOC
- CVAT XML
- Mask PNG
- Custom JSON
Tools we work in
- CVAT
- Label Studio
- Labelbox
- Encord
- V7
- Your in-house tool
Pricing
Bounding boxes from $0.035 per box, polygons from $0.08 per object and semantic segmentation from $0.60 per image, with volume rates from 50,000 units.
How we keep it accurate
Every project runs on the same four-check QA process. See how our QA works.
IoU checks on every batch
Box and mask overlap is measured against hidden gold items, so you see geometric accuracy, not just class accuracy.
Edge-case library
A living set of right-and-wrong examples for occlusion, reflections, truncation and lookalike classes.
Class-level reporting
Precision and recall per class, so a weak class is visible before it skews your model.
How a project runs
Scope and guidelines
We review a sample, agree classes, edge cases and acceptance criteria, and sign your NDA.
Pilot phase
Pilot phase helps you judge quality, speed and communication.
Certified team
Annotators pass a gold-set test before they are cleared for production.
Production and reporting
Batches with QA reports, in your format, on an agreed schedule.
Image annotation questions
What image annotation types do you offer?
Bounding boxes, polygons, semantic and instance segmentation, keypoints and pose, polylines and image classification, alone or combined in one workflow.
How accurate is your image annotation?
Standard projects target 95%+ against your gold set with two review layers; managed projects target 98% with three layers. The target and the metric (IoU, precision, recall) are written into your SOW.
Can you use model pre-labels?
Yes. If you have a model, we can correct its pre-labels instead of drawing from scratch, which usually lowers cost per image. Every pre-label is still reviewed by a person.
How many images can you label per week?
It depends on scene density and annotation type. We give you a realistic throughput figure after running a pilot, then scale the team in phases to meet it.
Send a sample. See the labeling quality first.
Share a few files and a line about your goals. A project lead replies within one business day with an approach, a timeline and a price.