Semantic & Instance Segmentation services · 03 / 10
Semantic & Instance Segmentation
Pixel-level masks: every pixel classed, every instance separated.
For scene understanding there is no shortcut: every pixel needs a class. We combine model-assisted pre-labelling with human refinement at the boundary, where automated masks fail most.
Typical use cases
- Autonomous driving scenes
- Surgical and pathology imagery
- Robotics manipulation
- Satellite land cover
Sample deliverable
What the output looks like.
Synthetic demonstration scene. No client data.
Variants
Semantic
Class per pixel across the full frame, with a void class for ambiguous regions.
Instance
Unique ID per object, so overlapping cars stay separate cars.
Panoptic
Stuff and things in one pass, delivered in standard panoptic format.
QA metric
Pixel mIoU vs gold
≥ 0.90
Reported per class so weak classes are visible, not averaged away.
Export formats
- PNG masks
- COCO RLE
- Cityscapes
- Panoptic JSON
- NumPy arrays
How every project runs
- Scope call & sample review
- Written guideline & gold set
- Fixed-fee pilot with QA report
- Production in weekly batches
Next step
Scope your project with
an annotation lead.
A 30-minute call with the person who would run your project. We review a sample of your data, recommend an approach and send a fixed-price proposal within two business days.