Unreliable Face Detection in Image Enhancement Software
-
AI image enhancement software cannot reliably identify pixelated or blurry faces in photos and videos. The results are inconsistent guesses rather than revealing truths.
-
When asked to "enhance" an image, AI programs add new information and details based on training data rather than uncovering what's really there.
-
AI functions best in specific cases, like enhancing blurry astronomical images, where training data provides accurate information to fill gaps. Faces have too much variation.
-
Some AI programs are tailored narrowly to generate a specific person's face, but still cannot determine if a blurry image actually depicts that person or someone else.
-
If a face is too pixelated or blurry for a human to confirm identity, there is not enough definitive data for an AI system to zoom in on and enhance accurately.