IQA

How it scores. Six print-quality dimensions, 0–100: sharpness (variance of Laplacian), exposure (brightness placement + clipping), contrast (tonal spread), noise (flat-region sigma), white balance (neutrality of bright pixels — true whites carry the cast), subject (skin-region presence — approximate in-browser; the on-prem engine uses real face detection). Weighted overall and grade match the Python iqa.py module, so browser spot-checks and batch CSVs line up with production logs. Images are analyzed locally and never leave this machine.