segmentation

---

name: segmentation description: Isolates skin regions from background clutter using calibrated thresholds for downstream analysis license: MIT compatibility: opencode metadata: audience: developers workflow: clinical-pipeline

What I do

I isolate skin regions from the background to focus analysis on the relevant areas. I use calibrated thresholds based on skin tone classification to create accurate skin masks.

When to use me

Use this when:

• Image preprocessing is complete and you need to segment skin regions
• You need to remove background noise for feature extraction
• You need a binary mask of skin vs non-skin pixels

Key Concepts

• Skin Mask: Binary image where skin pixels are 1, background is 0
• Calibrated Thresholds: Thresholds adjusted based on Fitzpatrick type
• segmentation_complete: State flag after segmentation complete
• Background Removal: Eliminate non-skin pixels for focused analysis

Source Files

• services/vision.ts: Segmentation implementation
• types.ts: WorldState interface

Code Patterns

• Apply calibrated color thresholds based on fitzpatrick_type
• Generate binary mask of skin regions
• Validate segmentation quality before marking complete

Operational Constraints

• Thresholds MUST be calibrated to Fitzpatrick type
• Poor segmentation affects all downstream accuracy
• Consider fallback to manual region selection if auto-segmentation fails