AI Initiative

Mapping the full spectrum
of human skin

A studio-based research project capturing ground-truth skintone data from 1,000 participants using spectrometers and the Monk Skintone Scale — building the most precise skintone dataset for AI.

Participate in the Study → Let Us Build Your Dataset
Project Overview

What we're building.

The most comprehensive, consent-based skintone dataset ever produced.

At our Newark studio, we're capturing precise skintone data from 1,000 participants across the full Monk Skintone Scale. Each participant completes a 30-minute session that includes spectrometer readings of their skin, and photography under two controlled lighting environments.

1,000
Participants
30 min
Per Session
2
Lighting Setups
$100
Amazon Gift Card
Studio Capture
Our Approach

Precision measurement meets cultural understanding

Unlike typical image datasets scraped from the web, every data point in this project is captured in a controlled studio environment with calibrated equipment. We use industrial-grade spectrometers to measure exact skin reflectance values — not subjective visual assessment.

Each participant is photographed under two distinct lighting setups: a neutral studio key light and a warm variable environment. This dual-capture approach ensures the data is useful for AI systems that encounter people in real-world lighting — not just lab conditions.

Trained cultural consultants guide every session, ensuring participants feel comfortable and represented. Consent is ongoing, not a one-time checkbox.

What We Do

Every session, every angle, every expression

During each 30-minute session, participants are photographed across multiple head positions and facial expressions — ensuring the dataset covers the full range of how faces appear in the real world. Each capture is paired with a spectrometer reading taken from the forehead, cheek, and dorsal hand — three sites that provide the most consistent skintone measurement across all populations.

Position 1Front
Position 2¾ Left
Position 3¾ Right
Position 4Profile Left
Position 5Profile Right
Position 6Neutral
Position 7Smiling
Position 8Serious
Position 9Eyes Closed
Position 10Looking Up
Position 11Looking Down
Position 12Tilted Left
Position 13Tilted Right
Position 14Mouth Open
Position 15Squinting
MST 1
1#f6ede4
MST 2
2#f3e7db
MST 3
3#f7d7b4
MST 4
4#eadaba
MST 5
5#d7bd96
MST 6
6#a07e56
MST 7
7#825c43
MST 8
8#604134
MST 9
9#3a312a
MST 10
10#292420
Skintone Data

The full Monk Skintone Scale, MST 1 through 10

Developed by Dr. Ellis Monk at Harvard University, the Monk Skintone Scale provides a more representative and inclusive framework for classifying human skin tones than legacy systems like the Fitzpatrick scale.

Our dataset targets balanced representation across all 10 MST values — ensuring AI systems trained on this data perform equally well for the lightest and darkest skin tones, and everyone in between.

Each participant's skintone is measured with a spectrometer and mapped to the Monk Scale — providing both objective colorimetric values and a standardized classification that AI systems can learn from.

Now Recruiting

Be part of the most inclusive skintone dataset ever created

We're looking for participants of all skin tones, ages, and backgrounds to join this study at our Newark, NJ studio. Your participation helps ensure that AI technology sees and understands everyone.

30-minute studio session at our Newark facility
Non-invasive spectrometer readings + photography
Two controlled lighting environments
Compensation provided for your time
Full informed consent — your data, your rights
Register to Participate →
Did You Know?

All cameras were calibrated from a single white woman named "Shirley"

For decades, Kodak's color film processing was calibrated using a reference image called a "Shirley Card" — a photo of a white woman with light skin, used as the standard for what "correct" skin color should look like in a photograph.

This meant that camera exposure, color balance, and print processing were all optimized for light skin tones. Darker skin tones were consistently underexposed, poorly rendered, and treated as edge cases rather than the norm.

When digital photography arrived, many of these same biases were carried forward into sensor algorithms and auto-exposure systems. And now, as AI systems are trained on decades of these biased images, the Shirley Card's legacy persists — AI models that struggle to accurately see, classify, and generate people with darker skin tones.

Our Skintone Research Project is a direct response to this history. By capturing precise, spectrometer-verified skintone data across the full Monk Scale, we're building the ground-truth dataset that finally replaces the Shirley Card's legacy with something that represents all of us.

Let us build your
custom datasets

Need bespoke visual data for your AI pipeline? Our studio produces rights-cleared, culturally annotated datasets tailored to your exact specifications.

Start a Project →