Define the 3×3 plus-sign image classifier data contract

Make malformed input fail before it reaches the interesting algorithm. The accepted contract is exactly three rows of three binary brightness values. This boundary matters because wrong dimensions, non-binary pixels, rotations, one-pixel corruption, and a threshold fitted only to memorized examples; allowing a bad value through makes later debugging look like an algorithm problem. Keep transformation functions separate from scoring or prediction so a test can identify which boundary changed the data.

The artifact's user-facing goal is specific: classify a tiny pixel grid as plus or other while keeping the pixels, score, threshold, and prediction visible. Its accepted input is exactly three rows of three binary brightness values. Those statements are intentionally narrower than “build an AI system.” Narrow scope lets us inspect every input and expected result, and it prevents a toy result from being presented as a production claim. The example implements or probes the input boundary. Copy it into a fresh file and run it without extra packages.

Save this as lesson.py and run python3 lesson.py. It uses only the language standard library, so the example is reproducible offline.

def validate(img):
 if len(img)!=3 or any(len(r)!=3 for r in img): raise ValueError('expected 3x3')
 if any(p not in (0,1) for r in img for p in r): raise ValueError('binary pixels only')
 return img
print(validate([[0,1,0],[1,1,1],[0,1,0]]))

Expected output: the validated 3×3 grid. Exact floating-point formatting may vary slightly, but the asserted behavior must not. Read the output as evidence about this stage, not merely proof that the interpreter started.

Print the 3×3 grid, flattened indices, five selected pixels, score, threshold, and label. If a rotated or shifted plus fails, that is an invariance limitation rather than a Python bug. Enumerate all nine single-pixel flips instead of trying one convenient corruption. Shape and binary-value errors should stop before scoring so malformed images cannot masquerade as confident predictions.

At the data contract stage, save the smallest failing fixture beside the expected result. Change one cause at a time and rerun the exact command printed above; that makes the repair reviewable and keeps this chapter's progressive artifact reproducible.

Use named positive and negative grids to compute confusion counts, then report every single-pixel-flip result. Accuracy alone hides whether the toy misses pluses or calls unrelated lines pluses, so include recall and false-positive count. Keep the threshold in the report and test a rotated plus separately. This evaluates a handcrafted detector, not a camera-ready vision model.

For this data contract page, preserve the fixture and result as evidence for the next page. Label observations separately from conclusions: a passing assertion establishes the behavior it names, while broader usefulness requires the chapter's full evaluation set and stated operating limits.