Childhood vs adult photos: a same-person stress test
A large age gap is a useful way to challenge a face-comparison system because the identity stays the same while facial proportions, camera quality and appearance can change dramatically.
The question
Can a face-comparison system still produce a sensible same-person result when the two photographs are separated by many years? This is harder than comparing two recent selfies. A child's face changes in proportion as the skull, jaw, nose and soft tissue develop, while hairstyle, camera quality and image processing may be completely different.
How we perform this check
We use known same-person photo pairs where the identity is not in doubt before the comparison begins. One image can come from early childhood and the other from the teenage or adult years. We then repeat the check with ordinary real-world differences such as glasses, headwear or a different crop.
| Variable | Why it matters | What we look for |
|---|---|---|
| Large age gap | Facial proportions change during growth. | The system should not rely only on age-specific surface appearance. |
| Glasses | Frames cover part of the eye region. | A known same-person result should remain sensible. |
| Headwear | Hairline and upper-head context may be hidden. | The comparison should still focus on the visible face. |
| Old vs modern camera | Resolution, contrast and lens behavior differ. | Image quality should not dominate the identity decision. |
What we observed in manual checks
Known same-person childhood-to-adult pairs have remained recognizable to the system even when the photographs were not visually identical and the newer image contained ordinary appearance changes such as glasses. This is an encouraging development check because it suggests the comparison is not based only on hairstyle, clothing or one narrow age-specific appearance.
That observation should not be turned into a universal accuracy number. A childhood photo can still be too small, blurred, strongly angled or partially covered for a useful comparison. The practical lesson is that a large age gap does not automatically make a comparison meaningless, but photo quality still matters.
Why a score can still change
Even for the same person, two image pairs can produce different similarity percentages. The model sees the pixels it receives. A front-facing school photo and a low-light mobile selfie are not equivalent inputs. This is why LooksCompare treats the percentage as an image-based similarity estimate rather than a permanent “identity score” attached to a person.
A better way to test an old photo
- Use the clearest old photo available and avoid a tiny face crop.
- Prefer a newer photo where the face is roughly frontal and not heavily filtered.
- If the first result surprises you, repeat the comparison with a second known photo rather than changing only one image until you obtain the answer you want.
- Judge consistency across several sensible photo pairs, not one percentage in isolation.