Skip to content
The Face Shape Detector
Face features

Eye Shape Detector

Measure canthal tilt, lid crease and eye spacing from a photo to place you among the ten eye shape descriptors.

Loading tool…
  • 468

    Landmarks mapped

  • 0

    Data collected

  • Free

    Always, no account

  • Any

    Device or browser

3 simple steps

How the eye shape detector

What happens from the moment you start, in order.

  1. 1

    Take a bare-faced eye photo

    Look straight into the lens at eye level, brows relaxed, glasses off. MediaPipe maps the lid ring around each eye in your browser, fixing both canthi, the lid margins and the pupil center.

  2. 2

    Measure tilt and aspect

    Fissure width divides by height to give the aspect ratio, canthal tilt is read in degrees against a roll-corrected horizontal, and your inner corner gap is compared against a single eye width.

  3. 3

    Read crease and orbital depth

    Crease detection separates monolid from hooded and creased lids, while lid curvature and shading gradient estimate whether the globe sits deep in the orbit or forward of the rim.

What you get

Built to be checked

Every reading comes with the measurement behind it.

Ten shapes, three families

Outline, lid and position traits get reported separately, so you can read as almond, hooded and close-set at once rather than being forced into one label.

Canthal tilt in degrees

See the measured angle between your inner and outer corners after head roll correction, alongside fissure aspect ratio and intercanthal spacing.

Nothing leaves your device

Landmark detection happens in the browser through Google MediaPipe. Your photo is never uploaded, never stored, and disappears the moment the page closes.

Technique matched to lids

Each result carries liner, shadow placement and brow guidance written for that lid structure, including where a cut crease belongs on a hooded eye.

Eye shape classification (palpebral morphology typing) describes the visible eye opening and its setting in the skull using ten descriptors drawn from three separate families. Outline traits are almond, round, upturned and downturned. Lid traits are monolid and hooded. Position traits are close-set, wide-set, deep-set and protruding.

Three ratios carry most of the decision. Fissure width over fissure height sits near 3.0 on an almond eye and under 2.5 on a round one. Canthal tilt, the angle from medial canthus to lateral canthus against a roll-corrected horizontal, reads upturned past two degrees positive and downturned past two degrees negative. Intercanthal spacing uses 1.0 as its reference, meaning the gap between inner corners equals one eye width; below 0.9 is close-set and above 1.1 is wide-set.

Google MediaPipe Face Mesh places the landmark ring these figures come from, 468 points across the whole face, computed inside your browser. Nothing is uploaded, nothing is stored, and no server ever sees the picture.

The sections below cover where the vocabulary came from, the five landmark points and the values derived from them, the ten descriptors compared in a single table, liner and shadow placement matched to each lid structure, the capture conditions that corrupt a small-feature measurement, and which parts of the output deserve your trust.

Where the vocabulary came from

Vocabulary for eyes was built by two professions that needed it daily. Makeup artists wanted to explain why one liner technique flattered a client and closed down the next one. Clinicians wanted terms for lid position and orbital depth.

Today's ten categories merge both vocabularies, which is why some name an outline and others name a setting. The merger has a consequence worth knowing before you read a result: the families are not mutually exclusive. You can be almond in outline, hooded in lid structure and wide-set in position at the same time, and most people are exactly that combination of three. Sorting them apart takes measurement.

Landmarks and the values derived from them

Landmarks used here run to five per eye: the medial canthus, the lateral canthus, the highest point of the upper lid margin, the lowest point of the lower lid margin, and the pupil center. Everything reported is derived from those five.

  • Aspect ratio divides the width of the palpebral fissure by its height at the tallest point, separating elongated openings from circular ones.
  • Canthal tilt measures the corner-to-corner line against horizontal once head roll has been removed from the mesh.
  • Intercanthal ratio compares the gap between the inner corners against the width of a single eye.
  • Crease visibility records whether a supratarsal crease exists above the lid margin, and whether skin above it overhangs that margin with the eye open.
  • Orbital projection estimates how far the corneal apex sits forward of the brow ridge and lower orbital rim, read from shading gradient and lid curvature.

Those five values map onto the ten descriptors in the table that follows.

The ten descriptors compared

DescriptorFamilyDefining signal
AlmondOutlineElongated opening, visible crease, iris clipped top and bottom
RoundOutlineCompact opening, full ring of white around the iris
UpturnedOutlineOuter corner sits above the inner corner, positive tilt
DownturnedOutlineOuter corner sits below the inner corner, negative tilt
MonolidLidNo supratarsal crease, one smooth plane from lash line to brow
HoodedLidCrease present, covered by an overhanging fold of upper lid skin
Close-setPositionInner corner gap narrower than a single eye width
Wide-setPositionInner corner gap broader than a single eye width
Deep-setPositionGlobe sits back under a prominent brow bone, naturally shadowed
ProtrudingPositionGlobe sits forward of the rim, lid reads convex and full

Almond serves as the reference outline, iris just touching both lid margins. Round shows sclera all the way around, which the eye reads as openness. Upturned and downturned are pure tilt and attach to any lid type. Monolid gives you a lash line as the only structural line to work with, while hooded hides its crease in the open position, which is why product placed on the lid disappears. The four positional descriptors say nothing about the opening itself, only where it sits, and each combination calls for different product placement.

Technique matched to lid structure

Technique follows one idea: put emphasis where you want the eye to look larger or shifted, and keep weight off whatever already dominates.

  • Almond eyes take nearly any approach; a gradient with the darkest tone in the outer third is the safe default.
  • Round eyes lengthen when liner extends past the outer corner instead of thickening all the way around.
  • Upturned eyes balance out when the lower lash line carries weight along its outer third.
  • Downturned eyes lift under a wing aimed at the brow tail, with lower liner stopping two-thirds along.
  • Monolid eyes need a wider horizontal gradient built up from the lash line, since pigment has no fold to sit in.
  • Hooded eyes keep their transition shade visible only when it goes above the natural crease, with liner thin and tight to the lashes.
  • Close-set eyes push apart when the inner corner is lightened and depth concentrates outward.
  • Wide-set eyes draw together the opposite way: definition carried inward, outer edge left soft.
  • Deep-set eyes open up under light reflective tones, with dark shadow kept out of an already shadowed crease.
  • Protruding eyes recede under matte mid-tones; shimmer pushes the lid further forward.

Every one of those readings assumes the photograph was taken properly, and eyes are the least forgiving feature on that count.

Capture conditions that decide the reading

Capture matters more here than for any other facial measurement, because the structures are small and lid position changes by the second.

Look straight at the lens, expression relaxed, both eyes open but not widened. Raising your brows lifts the upper lid and can hide a hooded fold completely. Keep the camera exactly level with your eyes. Shooting from below exposes sclera under the iris and can turn an almond reading round.

Light should be even and frontal. Overhead light drops shadow into the orbit and makes any eye look deep-set, while hard side light exaggerates brow overhang. Take the glasses off, clear hair and fringe from the brow line, and photograph bare eyes where you can, since liner exists precisely to alter apparent outline and tilt. Even a careful photo has boundaries, and it helps to know which numbers sit inside them.

What the numbers support and what they don't

Numbers hold up well for the first three values. Aspect ratio, tilt in degrees and inner-corner spacing repeat closely between photographs taken under matched conditions, and they're the trustworthy part of any result.

Labels hold up less well. Depth classifications are the weakest of the set, because orbital projection is a front-to-back quantity that a frontal photograph does not contain. The tool infers it from shading and curvature, and strong light produces false readings in either direction. Hooding is a matter of degree rather than a switch, and it increases gradually across adult life, so plenty of people sit legitimately in between.

None of this is a clinical assessment. Mild differences between your two eyes, small variations in crease height and gradual changes in overhanging skin are ordinary. But they aren't something a geometry tool should interpret. Concerns about vision, drooping or a change you've noticed recently belong with an optometrist or ophthalmologist.

Ten descriptors, three families, one browser tab. The vocabulary merges makeup practice with clinical terminology, the five landmark points per eye produce aspect ratio, canthal tilt and intercanthal spacing, and the comparison table shows the defining signal behind each label. Technique follows the lid structure rather than the outline alone, capture conditions decide whether a small-feature measurement means anything, and the depth-based labels carry the widest error bars of the set. Nothing here diagnoses. Open the eye shape detector, take a level bare-eyed photo, and read the degrees alongside the names.

Questions

Eye Shape Detector FAQ

The questions people ask most often about this tool.

What are the ten different eye shapes?

Almond, round, upturned, downturned, monolid, hooded, close-set, wide-set, deep-set and protruding. They split into three families: outline shape, lid structure and position in the orbit. Because the families describe different things, most people match one from each. Almond in outline, hooded in lid structure and wide-set in position is an ordinary combination rather than a contradiction.

What is canthal tilt?

Canthal tilt is the angle of the line drawn from the inner corner of the eye to the outer corner, measured against horizontal after head roll is removed. A positive angle puts the outer corner higher and produces an upturned eye. A negative angle puts it lower and produces a downturned one. Tilt decides where liner and shadow belong.

Is my photo uploaded when I use this tool?

No. The eye landmark mesh is computed by Google MediaPipe running locally in your browser tab, so the image is processed on your own hardware. Nothing is transmitted to a server, nothing is saved to a database, and the photo clears from memory as soon as you leave or refresh the page.

Can my two eyes be different shapes?

Yes, and mild differences are the norm rather than the exception. Small variations in tilt, crease height and how much lid skin overhangs turn up in most faces, and hooding often develops unevenly across adult life. Each eye is reported separately so the difference is visible. A sudden or noticeable change is worth raising with an optometrist.

Which is more reliable, canthal tilt or a deep-set reading?

Canthal tilt wins on repeatability. It is an angle between two sharply defined landmarks, so it holds steady across photos taken under matched conditions. Orbital depth is a front-to-back quantity that a frontal photograph does not contain, so the tool infers it from shading and lid curvature. Overhead light alone can produce a false deep-set reading.

How can I tell if I have hooded eyes?

Look straight ahead in a mirror with your brows completely relaxed. If a fold of upper lid skin covers the crease so that little or no lid space shows between lash line and fold, your eyes are hooded. Raising the brows hides the effect, which is why many people notice hooding in photographs long before they see it in a mirror.

Start with your face shape

Most styling decisions follow from the shape of your face. Run the detector, then work outward from the result.