Beard or no beard? Test one thing at a time.
Shaving or taking off your glasses changes the image the model reads. To know what it really changes, everything else has to stay the same.
Updated 2026-10-18
What research says about beards
In a study published in 2008, 60 women rated computer-edited male faces shown clean-shaven, with stubble, a light beard or a full beard. Full beards were judged the oldest and the most mature. People, not algorithms, gave these ratings, but apparent-age models learn from human judgements: FairFace, the model used here, was trained on ages estimated by annotators.
NIST makes a related point about algorithms. Looking at years of daily selfies of the same man, it notes that he grew a heavier beard over time, and that this is a choice, not part of ageing. A model can still react to it.
What research says about glasses
NIST compared photos of the same people with and without glasses: 9,181 people aged 14 to 36, 16,677 pairs. The effect depended on the algorithm. Some estimated people as older with glasses, others estimated them younger. There is no general rule, so the only way to know is to test both versions under the same conditions.
Glasses also cause practical problems: reflections that hide the eyes, thick frames that cover the brows, tinted lenses. The face detector needs to see the eyes clearly to align the face.
The method: one change per pair of photos
- Same session. Both photos within a few minutes, same place, facing the same window. Light alone can change the age group for about one face in six.
- Same framing. Same distance, camera at eye level, same head angle. Mark your position on the floor if needed.
- Same expression. Neutral face in both photos. A smile changes the reading more than you think.
- One change only. Glasses on, then off. Or beard, then after shaving the same day, with the same light.
- Check the conditions. Put both photos into the photo comparison tool: it flags differences in brightness, sharpness and framing before you read the result.
Reading the result honestly
If both photos land in the same age group, the change made no visible difference for the model. If they land in two neighbouring groups, it may be the change, or a face close to a group boundary: repeat with a third pair before concluding. Face-Age has not yet measured the effect of glasses or beards on its own model; the limits guide lists what is still unmeasured.
Sources
- Neave and Shields, The effects of facial hair manipulation on female perceptions of attractiveness, masculinity, and dominance in male faces, Personality and Individual Differences 45(5), 2008.
- NIST IR 8525, Face Analysis Technology Evaluation: Age Estimation and Verification, Hanaoka, Grother et al., May 2024.
- Kärkkäinen and Joo, FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age, arXiv:1908.04913.