The classroom problem
Students in hands-on science and electronics activities work with breadboards, resistors, LEDs, wires, sensors, and small programmable boards. A single misplaced connection may prevent an experiment from working. The student sees “nothing happened,” but does not necessarily know whether the fault is in the wiring or their program.
The long-term goal is a system that can give useful feedback: which connection needs attention, where it is, and what to do next. That is more useful than a bare “incorrect” label, and it can reduce the amount of one-to-one troubleshooting required of teachers.
Why a reference photo won’t solve it
The same working circuit can be built in different physical locations. A resistor connected in one column can be moved, with the other connections moved appropriately, and the circuit can remain electrically equivalent. A photo comparison would see two different images; a circuit checker should see the same connectivity.
Three things make the problem different from ordinary image classification:
- Position has electrical meaning. Adjacent holes can belong to different electrical nodes. Being one column off may matter more than correctly identifying the component’s color.
- There can be many correct layouts. The software needs to reason about the electrical graph, not memorize one expected arrangement.
- Some faults are invisible. An LED might be reversed or defective even if the photograph looks normal. Hardware measurements help, and the system must admit uncertainty.
A small change, a real failure
A typical breadboard contains rows of holes internally connected in groups. Five holes in a strip share an electrical node, but the next column is a different node.
| Observation | Electrical consequence |
|---|---|
Wire goes into a32 |
Connected to the node containing a32 through e32 |
Wire moves to a33 |
Connected to a different node |
| Two parts occupy electrically equivalent holes | May represent the same circuit despite visually different positions |
| LED appears normal from above | Orientation or actual function may still be unknown |
Therefore the real question is not “Does this picture match the answer?” It is “What electrical connections does this photograph imply, and do those connections satisfy the lab?”
The intended student experience
When the system becomes reliable enough, the student should get one of three outcomes:
- Working circuit: Wiring has been sufficiently verified; the student can investigate their program or another cause.
- Actionable fault: A specific connection or component needs changing, ideally highlighted on the original photo.
- Uncertain: The evidence is inadequate, so the student is asked for another view rather than given a confident but invented correction.
The third outcome is an important part of correctness. A false reassurance that the wiring works could waste more time than admitting that the image is unclear.
Our contribution
The inherited software already compares circuits represented as structured data. Our primary work is translating visual evidence into that data without guessing beyond what the image supports.
Read the engineering pipeline or go directly to the computer-vision problem.