Accepted inputs
- Images and extracted video frames
- Raster mask proposals
- Polygon pre-annotations
- Client-hosted platform tasks
PIXEL-LEVEL ANNOTATION
Semantic and Instance Segmentation Services
Segmentation quality depends on explicit boundary rules. We align how to treat holes, thin structures, overlaps, truncation and uncertain edges before production, then review masks at the object and class levels.
SCOPE
What this service covers
Semantic segmentationInstance segmentationPolygon and multipolygon masksHoles and disconnected regionsBackground and ignore regionsMask correction and refinement
Annotation approach
- Pixel-class and instance-ID mapping
- Boundary inclusion rules
- Occlusion ordering
- Minimum-region thresholds
- Mask-to-source dimension validation
Common edge cases
- Thin or low-contrast boundaries
- Overlapping instances
- Interior holes
- Motion blur and compression artifacts
- Partially visible damage or tissue regions
QUALITY ASSURANCE
Review that follows the guideline
Quality checks are selected for the data type and acceptance criteria. Metrics and review coverage depend on the project, platform and agreed workflow.
Boundary adherence checksClass and instance separation reviewMissing-region and spill detectionMask dimension and encoding validationPolygon closure and geometry checks
Applications
- Vehicle damage
- Medical and pathology imaging
- Autonomous perception
- Aerial mapping
- Industrial inspection
- Agricultural vision
Output formats
- COCO polygons or RLE
- PNG masks
- CVAT XML/JSON
- LabelMe-style JSON
- Client-defined mask structures
What to include in your brief
- Class map and priority order
- Boundary and hole rules
- Treatment of hidden regions
- Minimum area or visibility threshold
- Required mask encoding and QA criteria
START WITH A CALIBRATION BATCH
Reduce uncertainty before production.
Share your guidelines, sample characteristics, volume, preferred platform and timeline. We’ll assess the scope and propose a representative calibration step.
Discuss your dataset