How PCBMind works

The inspection pipeline

Every uploaded photo moves through the same stages before you see a result:

1. 1Upload — your photo is stored and queued for analysis.
2. 2Image quality check — basic validation that the image decoded correctly.
3. 3Preprocessing — resized and contrast-corrected (CLAHE) so uneven lighting doesn't hide defects.
4. 4AI detection — the trained model scans the board for defects. This is the slow step — a cold server can take up to a minute.
5. 5Component matching — detections are compared against your golden PCB reference, if one's attached.
6. 6Defect analysis — each detection is classified by type and severity.
7. 7Report — a PDF report is generated on demand from the results.

Roles & permissions

Every user belongs to an organization with one of three roles:

The first person to sign up for an organization becomes its admin.

Defect taxonomy & severity

Every detection falls into one of six defect types, each mapped to a severity level:

API basics

The frontend talks to a REST API secured with a Supabase-issued bearer token — every request carries Authorization: Bearer <token>. Full interactive API docs (OpenAPI) are available at /docs on the API host.

FAQ