From two photos to 13 body measurements, in seconds.
Your user takes one front photo and one side photo with their phone, and enters their height and weight.
The API returns 13 measurements in centimeters, including chest, waist, hips, and shoulders.
No tape measure, no setup, a single HTTP request.
Under the hood, silhouette extraction, height/weight planes, and an MNASNet-based regression network do the work — all you see on your side is a single POST.
01
Send the photos
Front and side photo, JPEG or PNG, up to 8 MB. Plus height and weight. multipart/form-data, nothing else needed.
Good example
FrontSide
02
Silhouette + model
The background is removed, the two silhouettes are merged, and height and weight are added as separate channels. The model produces 13 values in a single pass.
03
Get the measurements
13 measurements in centimeters come back as JSON, along with short field names you can write straight into your user record.
Live demo
Try it with your own photos.
The form below sends a request through the secure demo proxy. Photos are never stored on the server — they're dropped from memory as soon as the prediction finishes.
For best results: the person should be fully visible head to toe, the background should be plain and high-contrast,
and the side photo should be taken turned 90°. Loose clothing inflates girth measurements.
Upload both photos, then click "Predict measurements".
Use cases
Any product that needs body measurements.
Healthcare
Automate measurement collection for waist-to-hip ratio, body composition, and risk scores.
Waist/hip ratio tracking
Remote patient assessment
Personalized nutrition plans
Sports & fitness
Measure user progress by actual girth measurements, not just weight.
Weekly measurement change charts
Muscle/girth growth tracking
Adaptive training programs
Apparel & e-commerce
Accurate size recommendations are the fastest lever for cutting return rates.
Automatic size-chart matching
Return rate reduction
Measurements for made-to-order production
1.82%average error margin — roughly ±1.5 cm at 175 cm height
13measurements: chest, waist, hip, shoulder, arm, leg, thigh, calf, and more
2.5–6 secsingle prediction time on CPU; no GPU required
Statelessno database, no photo storage, scales horizontally
Body measurement from two photos is still a rare feature. Shipping it puts something in your product that nobody else in your category offers — and it works accurately enough to build on.
Low error margin
1.82% average error — around ±1.5 cm on a 175 cm person. Close enough that size recommendations, progress tracking, and health ratios hold up in production instead of staying a demo.
Trained and validated on real body scan data
13 measurements, each with consistent accuracy
No manual tape measure, no user self-reporting errors
Fast response time
2.5–6 seconds per prediction. Your user uploads two photos and sees results before they lose interest — no queue, no "we'll email you when it's ready".
Predictable cost per prediction
Stateless service, scales horizontally under load
Fits inside a normal request/response flow
Still a rare feature
Almost nobody offers this. Most apparel, fitness, and health products still ask users to type in their own measurements — and most of those numbers are wrong or never entered at all.
A real differentiator, not another checkbox feature
Removes the friction step users abandon
Early enough that being first still counts
Something new in your service
The measurements become a foundation you can build on: sizing that actually fits, progress charts that mean something, and personalized plans driven by real numbers rather than guesses.
Lower return rates for apparel and e-commerce
Body-based progress instead of weight alone
New premium tiers you couldn't offer before
Want to try it in your product?
Tell us about your usage volume and integration scenario; we'll put together a test key and a sample integration for you.
