I have a small home electronics lab and I like prototyping projects for home automation, guitar pedals and other “toy projects”. A lot of time goes into visually double-checking wiring, component values, and orientation (is that diode really in the right way round, is that resistor band really 4.7k and not 47k), small mistakes that are easy to make and annoying to debug after the fact.
I want to use Voyager Wingman to build a camera-based "bench companion": point a camera at the breadboard/PCB, and get real-time identification of components (resistor color-code values, capacitor markings, IC package/pinout) plus basic wiring sanity checks against a reference schematic.
Another use case is reverse-engineering a pcb to add new functionalities to it, for example now I'm working on making a “smart” dehumidifier from a “dumb” one, and an automatic real-time recognition of components and wiring could come in handy.
The first easier task would be to recognize standard pass-through components, their values and polarities, perhaps moving around the board or the camera to get different point of views.
SMD components could come next, together with pcb wiring (where visible) and silkscreen markings.
Another improvement would be recognizing issues in soldering, burnt components, blown caps and so on.
Why this project
- Genuinely useful, solve recurring problems in prototyping.
- Good showcase for Voyager Wingman on a vision task that's more about recognition and classification than raw object detection: component identification, text/marking reading, layout comparison.
- Natural extension path during the build weeks: start with simple component recognition, then add schematic-vs-breadboard comparison and other bonus improvements.
