Muggie is an artificial intelligence program that has been trained to recognize an attractive male and female face. Unlike other programs, Muggie is not scoring based on facial symmetry, golden ratios, or other measurements. We literally trained Muggie by showing it photos of attractive and unattractive faces, and it learned over time to recognize an attractive face. To eliminate any bias, we used models and celebrities who are widely considered attractive for the training set, which includes a variety of ethnicities from around the world.
Here are some things to keep in mind when using the app:
- Faces do not need to be perfectly front facing. In fact, off angle photos, tilted images, etc will work just fine.
- We trained Muggie to work with both smiling and neutral expressions.
- Bangs and hairstyles that hang down on the face are okay for the most part. Some high scoring female images we’ve tested have hair covering half of the face.
- Zooming does make a difference. Zooming too far out or too close will lower the probability score. The face should be around 80%-90% of the screen.
- The female predictor is working much better than the male predictor at this time. There’s just more variation with males, e.g. facial hair.
- Muggie is not analyzing the person, only the photo. An attractive person can take a bad photo and that will be reflected in the score.
- Poor lighting, blurry, or grainy photos will not score well even if the person is attractive. It’s rating the face as well as the quality of the photo.
- Use color photos only. Black and white images don’t score well as the program was trained with color photos.
- Muggie does differentiate between makeup and no makeup for women and will give higher scores for images with makeup. However too much makeup will generate a lower score.
We aren’t finished! We will continue to train the program with new images and create new releases.