Mandinka skin tone: Fitzpatrick type and undertone in 47 observed portraits
Mandinka: fitzpatrick skin type VI in 94% of 47; undertone warm in 79% of 47.
Catalog portraits for Mandinka are AI-generated adult subjects; the measurements on this page come from documented photographs of real people, not from these images.
Fitzpatrick skin type
n = 47 assessableSix-class ordinal classification of constitutive skin pigmentation, originally developed for sun-reactivity assessment but now the de-facto skin-tone vocabulary in dermatology and computer-vision fairness research.
| Value | Share | n | |
|---|---|---|---|
| IVBurns minimally, tans easily and well. Olive to light brown skin. Common in Mediterranean, Middle Eastern, Latin American, Southeast Asian populations. | 2% | 1 | |
| VRarely burns, tans deeply. Brown skin tone. Common in South Asian, North African, Middle Eastern, some Latin American populations. | 4% | 2 | |
| VINever burns, deeply pigmented at baseline. Common in Sub-Saharan African, African-diaspora, some South Asian, some Pacific Islander populations. | 94% | 44 |
Vocabulary: Fitzpatrick skin type (fitzpatrick_1988)
Undertone
n = 47 assessableHue character of the skin independent of overall depth (Fitzpatrick). Reflects relative balance of hemoglobin (red), melanin (brown), and carotenoid (yellow) contributions to perceived color.
| Value | Share | n | |
|---|---|---|---|
| warmYellow, peach, or golden cast. Veins on inner wrist appear greenish. | 79% | 37 | |
| coolPink, rosy, or bluish cast. Veins on inner wrist appear blue or purple. | 21% | 10 |
Vocabulary: Skin undertone (undertone_qualitative)
Where these numbers come from
47 publicly documented portraits of notable Mandinka people were assessed by a vision model against the catalog vocabulary, one row per portrait. Values are what the photographs show, read by a model, not population measurements. Each table counts only portraits where the value could be assessed; covered, obscured and unclear portraits are reported separately.
Notable-people corpora skew older, so hair color over-represents gray and white, and image quality varies. Treat shares as a description of this sample. Mean model confidence for this family: 0.82. 51% of portraits are high quality.
The same rows are published as image_observations in the catalog dataset; see the methodology. A family page exists only when at least 10 portraits carry an assessable value.