PCB-Labeller-Demo

May 4, 2026 · View on GitHub

Proof of concept: using AI to annotate PCB (and other hardware) images and help the user plan a project, coupled with light research assistance.

Idea

Point a vision-capable model at a photo of a PCB (or breadboard, module, or other piece of hardware) and have it:

  1. Identify visible components, pin headers, ICs, connectors, silkscreen markings.
  2. Produce an annotated overlay (boxes / labels / callouts) on the original image.
  3. Suggest next steps for the user's stated project goal — wiring, missing parts, datasheets to look up, likely pitfalls.
  4. Pull in reference material (datasheets, pinouts, similar projects) to ground the suggestions.

Status

POC / scratchpad. Not a finished tool.

Planned pieces

  • inputs/ — sample PCB photos to test against
  • outputs/ — annotated images + per-image notes
  • prompts/ — prompt templates for the labelling + planning passes
  • scripts/ — glue code (image annotation, research lookups)

Why

Hands-on hardware work benefits from a "second pair of eyes" that can read silkscreen, recognise common parts, and surface relevant docs without breaking the user's flow.