Facial geometry refers to the mathematical description of a face in terms of the spatial relationships between facial landmarks, eye corners, nose tip, mouth corners, jaw shape, and dozens of other measurement points. Machine learning models can generate compact numerical representations (face embeddings) that capture this geometry and allow rapid comparison across billions of stored face representations. ScanErase's biometric scan generates a face embedding from the uploaded photo and compares it against 2.4 billion indexed face embeddings across the web, finding deepfake and authentic NCII content by facial geometry match, not by pixel similarity.

Key facts about this term

  1. Upload a clear, front-facing photo for best scan accuracy ScanErase's facial geometry analysis works best with clear, front-facing photos with good lighting.
  2. Facial geometry matching identifies deepfakes that look different AI-generated deepfakes preserve facial geometry even when altering the body and background, this is what the scan matches.
  3. Higher quality photos yield more precise geometry embeddings A high-quality reference photo maximizes the accuracy of the biometric scan.

Frequently asked questions

Why can ScanErase find deepfakes when the images look nothing like my real photos?

AI tools preserve the victim's facial structure (facial geometry) when generating deepfakes because the face is the identifying element. ScanErase matches geometry, not pixels, finding content regardless of how the rest of the image has been altered.

Is facial geometry matching the same as facial recognition?

They are related but distinct. Facial recognition typically identifies individuals by name from live images. ScanErase's facial geometry matching identifies content hosting your facial structure regardless of other image characteristics.

Related glossary terms