When you upload a photo to the Doppelgänger Finder, the system doesn't just look at the photo as a whole. Instead, it translates your face into a series of numbers and coordinates. This mapping process relies on a computer vision concept called **Facial Landmark Point Detection**.
The Standard 68-Point Landmark Model
Most face recognition engines use a standard model that identifies 68 specific points (or nodes) across the human face. These coordinates are distributed as follows:
- Jawline Outline: Points 1 to 17 map the curvature of the jaw and chin.
- Eyebrows: Points 18 to 22 map the left eyebrow; 23 to 27 map the right eyebrow.
- Nose Structure: Points 28 to 36 map the bridge, nostrils, and tip of the nose.
- Eyes Geometry: Points 37 to 48 map the outline of both eyes, calculating width and height.
- Mouth Coordinates: Points 49 to 68 map the outer and inner contours of the lips.
Converting Points to Math
Once these 68 points are detected, they are treated as a geometric graph. The AI calculates angles and vector distances. For example, it calculates the ratio of the distance between your eyes relative to the width of your mouth.
By focusing on ratios rather than absolute pixel values, the scanner remains extremely accurate even if one photo is taken closer to the camera than another. It measures the pure proportions of your head.
Why Symmetrical Alignment Matters
When comparing two faces, the algorithm aligns both grids together and calculates the offset between corresponding points. If the offset is close to zero, it indicates that the two faces have identical bone structure, resulting in a high similarity match score.
