Can AI Really Measure Your Ring Size?

September 2, 2026

I have to admit, I sometimes give my Shop Assistant some pretty strange assignments.

This particular one started with a question that seemed simple enough:

Could AI actually determine someone's ring size from a photograph?

Not just look at a hand and make a wild guess, but use an object of known size in the photograph as a reference, do some actual visual measuring, and come up with a reasonably accurate ring-size estimate.

As a jewellery maker, this immediately caught my attention.

Because if this worked, even reasonably well, it could potentially be a very useful tool for people who shop for rings online but don't know their size.

So, naturally, I decided to make Allison, my assistant, the guinea pig.

🤖 The Canadian quarter experiment

The first thing I realized was that AI needed a reference.

A photograph by itself doesn't tell you the actual physical size of anything.

AI can see a finger, but it can't automatically know whether that finger is 16 mm, 18 mm or 22 mm across.

So I thought of something with a known measurement:

A Canadian quarter.

A Canadian quarter has a diameter of 23.88 mm.

That gave Allison a physical reference point.

The basic idea was:

Known object → establish scale → measure the finger visually → estimate circumference → convert to ring size.

Sounds easy, right?

Well… the first attempt taught us that how you take the photograph matters A LOT.

📸 Attempt #1: close… but not quite

For my first photograph, Allison put the Canadian quarter beside her fingers and asked AI to evaluate her ring finger size.

She received an estimate of approximately:

Size 7½

Not terrible.

Except she already knows her actual ring size.

Her finger circumference is approximately:

2.3 inches / 58.4 mm

That corresponds to approximately a US/Canadian size 8½.

So the first attempt was off by about one full ring size.

And that's where I started getting really interested.

Because instead of saying "Well, AI can't do this," I thought:

What if we give it a better photograph?

📸 Attempt #2: Let's make the photograph better

For the second test, she changed the setup.

She laid her hand flatter, made sure the relevant finger was clearly visible, and positioned the Canadian quarter beside the hand as a scale reference.

This time, Allison had AI estimate her ring finger at:

Approximately size 8½.

And that's her known actual ring size.

🎉

Okay.

NOW we're getting somewhere.

The second photograph produced an estimate that matched her known size.

Does that mean AI has officially replaced my ring mandrels?

Absolutely not.

😂

But it does demonstrate something pretty darn interesting:

The quality and composition of the photograph can make a significant difference to the usefulness of an AI-based measurement.

🤖 What is AI actually doing?

AI isn't magically "seeing" her ring size.

It's using visual proportions.

The Canadian quarter provides a known measurement of 23.88 mm. By comparing the apparent size of the quarter with the apparent dimensions of the finger, AI can estimate the finger's physical dimensions.

From there, it can estimate the circumference and translate that into a standard ring size.

The problem is that a finger is three-dimensional, while the photograph is two-dimensional.

That's where things get tricky.

Your finger isn't a perfect circle.

It isn't even necessarily a perfect oval.

And ring sizing depends on the circumference where the ring will actually sit, not simply the visible width of your finger in a photograph.

So there are plenty of opportunities for errors.

⚠️ Want to try this yourself? Here's where you can mess it up!

If AI-assisted ring sizing is going to be useful, the photograph needs to be taken properly.

🪙 1. Use a known reference object

This is essential.

A Canadian quarter works because we know its exact diameter.

Tell the AI exactly what you're using:

"This is a Canadian quarter. It is 23.88 mm in diameter."

Don't just say "here's a coin."

Different coins have different dimensions!

📏 2. Put the reference right beside your finger

Don't put the quarter on the other side of the table.

The closer the reference object is to the finger, the better.

Ideally, the quarter and your hand should be resting on the same flat surface.

📱 3. Take the photograph from above

This is probably one of the most important lessons from my experiment.

A photograph taken at an angle introduces perspective distortion.

For measurement purposes, we're not looking for artistic photography.

We're looking for:

"Security camera footage of my hand." 😂

Straight down is better.

☝️ 4. Keep the finger straight and relaxed

Don't curl your fingers.

Don't squeeze them together.

Don't stretch them unnaturally.

Just lay your hand naturally and make sure the finger you're sizing is clearly visible.

🔍 5. Show the whole finger

The AI needs to see where the ring will actually sit.

Don't crop out the base of the finger.

And make sure the entire relevant finger is visible in the photograph.

📸 6. Avoid extreme camera settings

Wide-angle lenses and very close photographs can create distortion.

A normal camera setting, with the phone held reasonably far above the hand, is preferable.

You want a measurement photograph—not a dramatic close-up of your fingerprints. 😂

💍 Why this could actually be useful

Here's why I find this experiment so exciting.

Imagine you're shopping for a ring online.

You see something you absolutely LOVE.

But there's one problem:

"What the heck is my ring size?"

You don't own a ring sizer.

You don't have a ruler handy.

There's no jewellery store nearby.

But you probably have a phone.

And maybe you have a Canadian quarter.

That's potentially enough to give AI a starting estimate.

And for online jewellery shopping, that's valuable.

It could potentially take someone from:

"I have absolutely no idea what size I am."

to:

"AI estimates I'm around an 8½, so now I know what size I need to verify."

That's a much more useful starting point.

🧪 Let's be VERY clear about the limitations

As cool as this experiment was, I wouldn't recommend treating an AI estimate as an absolute measurement.

There are too many variables.

Fingers vary in shape.

Two people can have similar-looking fingers but different circumferences.

Knuckles vary.

Some people have a large knuckle that needs to pass through the ring, while the base of the finger is significantly smaller.

Fingers swell and shrink.

Temperature, time of day and activity can all affect how a ring fits.

Wide rings fit differently.

A wide-band ring can feel tighter than a narrow band of the same nominal size.

And photographs aren't measuring instruments.

Even though AI can perform surprisingly sophisticated visual analysis, it is still working from an image.

So I would describe this as:

AI-assisted ring-size estimation

rather than:

Professional ring sizing.

🔬 The really interesting part of my experiment

For me, the most exciting discovery wasn't simply that Allison eventually got her size right using AI.

It was that changing the photograph changed the quality of the estimate.

First photograph:

AI estimate: approximately 7½

Improved photograph:

AI estimate: approximately 8½

Known actual size:

approximately 8½

That tells me that there is something worth investigating here.

If better photography produces better results, then perhaps we can develop a very specific set of instructions for people to follow.

And if those instructions consistently produce measurements within half a size?

Now we're talking.

🧠 What I'd test next

If I were going to turn this into a proper Le Dragon Argenté experiment, I'd want to test it on lots of different hands.

I'd use people with known ring sizes and photograph them using the same instructions.

I'd want:

Small fingers
Larger fingers
Different finger shapes
Prominent knuckles
Different ring sizes
Different lighting conditions
Different phones
Different hand positions
Different reference-object placements

Then we'd compare the AI estimates with the actual measurements.

That would tell us whether this is genuinely reliable—or whether my Canadian quarter has simply become a very entertaining jewellery laboratory assistant. 😂