I am writing a small test plan for a portrait upload form that includes face shape labels. The first question is whether changing the crop alters the label without changing the face itself. I would begin with a single authorized image and record the dimensions before making any copies. The copies would change only the surrounding space, leaving the face at its original proportions. I am considering <a href=https://faceshapecalculator.top>face shape detector as a reference while drafting the questions. These are proposed checks rather than results, and I have no consistency figures to share. For each copy, I could note its border dimensions, category and any confidence information provided. Would you compare only the final categories, or also record where the outline was detected? The acceptance rule should acknowledge ambiguity while still making unexplained changes easy to investigate.