Validate vision boxes against pixels before trusting coordinate labels

Vision-generated box coordinates need visual calibration against the exact decoded working image, even when the response asserts a pixel coordinate system and repeats the correct dimensions. Geometry checks prove bounds and area, not target coverage. Preserve the provider's original coordinates and receipt, independently inspect the overlay, and record corrected native coordinates with the correction reason. Do not apply one blanket normalized-to-pixel conversion when failures can be mixed. This applies to proposed image annotations; it does not establish human gold or training eligibility.