AI may be advancing fast, but it’s still only as good as what it’s trained on. In 3D morphological modeling, an algorithm is never better than the 3D dataset that trains it. That’s the observation that pushed MyFit Solutions to go well beyond its software platform, to also become a player in the acquisition and structuring of 3D data usable for R&D.
In short
A 3D dataset usable in R&D rests on three pillars: a rigorous, documented 3D data acquisition methodology, scientifically validated precision (ISO 5725-1:2023, roughly 99% scale accuracy), and the right technology choice for the need — smartphone for scalability, dedicated scanner for ultra-precision. MyFit Solutions offers both off-the-shelf datasets (like the Heads 800) and custom acquisitions.
In this article, discover the 3D data acquisition methodology that guarantees the reliability of our datasets, from initial capture through to integration into your R&D pipeline.
A “clean” 3D dataset doesn’t happen by chance
A usable 3D dataset is more than a pile of scans. For a piece of data to serve an R&D project, it has to follow a 3D data acquisition methodology that is precise, reproducible, and documented. Without that, the AI applied downstream has little chance of producing reliable results.
Capturing a human surface without error
Our teams and our partners carry out 3D data acquisition at scale, all over the world. That means mastering technical constraints specific to the human body: on areas that are too shiny or too dark, for example, adding texture or mattifying patterns avoids the holes that would otherwise appear in the reconstruction.
This protocol draws on the recommendations of ISO 20685-1, which governs scanning methodologies for internationally comparable anthropometric databases — a foundational reference for any serious 3D scanning acquisition protocol.
The real work starts after the scan
Once the data has been acquired, it still has to be prepared before it can be fed into an analysis pipeline. This normalization step is often underestimated, even though it can take a long time:
- Data cleaning
- Centering
- Scale normalization (scaling / rescaling)
- Feature normalization
- Data augmentation, depending on the use case
Discover our data offers
Scientifically validated precision
Claiming a 3D reconstruction is precise is easy. Measuring it, with an independent methodology, is harder. Our anthropometric data acquisition relies on a reconstruction engine that underwent a dimensional validation study conducted according to the ISO 5725-1:2023 standard, which distinguishes precision (repeatability) from trueness (closeness to physical reality).
The study compared our models to those of a reference 3D scanner used in the medical field, on a panel of 17 volunteers. Among them were healthy morphologies as well as patients with lymphedema (stages I to III), complex dysmorphic cases that are particularly demanding for a reconstruction algorithm.
An average coefficient of variation of 1.1% for repeatability. An average relative deviation of 1.1% compared to the reference scanner. That’s a scale accuracy of around 99%, including on pathological morphologies.
No statistically significant difference was found between pathological and healthy limbs (Wilcoxon test). Another key point: all of our meshes are fully manifold (watertight), a necessary condition for any downstream volume calculation or physical simulation.
Scalability or ultra-precision: two complementary approaches
The choice of technology for 3D scan acquisition for R&D depends first on the project’s goal. MyFit Solutions offers two approaches, which can also be combined depending on the phase of an R&D project.
Smartphone: unprecedented scalability
Our advanced photogrammetry technology turns the rear camera of a standard smartphone into a 3D capture tool. Anyone, anywhere, can collect reliable morphological data without expensive equipment. Behind the scenes, a serverless cloud infrastructure absorbs load spikes and makes it possible to generate several thousand models within a few months.
“It took us 6 months to acquire 500 precise hand scans, with 30 measurements extracted per hand on top of that. With our high-end 3D HD scanner, which costs tens of thousands of euros, it would have taken us 2 years, with a much higher risk of error.”
– Research project manager, medical device manufacturer
This capture method is particularly well suited to fitting studies for orthopedic medical devices, or to ergonomics studies for consumer equipment such as headphones. Our database includes several hundred hand scans covering a range of morphologies and ethnic origins, a useful foundation whenever a project calls for broad demographic representativeness.
3D scanner: high precision for R&D
For needs that require very fine anatomical detail, such as visible veins or micro-relief on the skin, we use dedicated 3D scanners. This is the right method for building ultra-detailed reference models, particularly ahead of scaling up via smartphone.
Discover our data offers
Smartphone or dedicated scanner: which approach should you take for acquiring your R&D 3D dataset?
| Criteria | Smartphone photogrammetry | Dedicated 3D scanner |
| Scalability | High, several thousand models within a few months | Limited by unit capture time |
| Cost per acquisition | Low | High (equipment costing tens of thousands of euros) |
| Level of detail | Suited to fitting and ergonomics | Ultra-detailed (veins, skin micro-relief) |
| Typical use case | Fitting studies, broad demographic representativeness | Reference models, validation ahead of scaling |
Summary table comparing the two acquisition approaches offered by MyFit Solutions.
Product focus: our 3D head dataset, Heads 800 dataset, ready to use
We’ve put together a specialized 3D head dataset of 800 head scans, designed to remove the technological barriers in the wearables and audio device sectors.
A protocol designed for the ear
Each subject was scanned wearing a silicone cap that compresses the hair, to reveal the true geometry of the skull. The ears remain fully uncovered, which allows for an accurate analysis of ear morphology, paired with the exact proportions of the head.
Two concrete use cases
On one hand, ergonomics and design: analyzing morphological fit to optimize the comfort of headphones, glasses, or protective helmets. On the other, acoustic simulation: modeling custom HRTF (Head-Related Transfer Function) filters, for spatialized sound rendering tailored to each user’s morphology.
What’s in the dataset
Complete demographic metadata, including gender, ethnicity, height, and weight, delivered together with the meshes in .PLY or .OBJ format.
MyFit Solutions carries out more than 100,000 3D scans and digital measurements every year, used daily in more than 15 countries, mainly in the health sector (medical devices, MedTech) and retail. That’s what turns an everyday smartphone into a genuine tool for capturing medical and industrial data.
FAQ
It relies on a documented, reproducible protocol (positioning, handling of reflective areas, post-processing), aligned with recognized standards such as ISO 20685-1, and scientifically validated.
It depends on the need: smartphones offer unprecedented scalability for large volumes, while dedicated scanners remain the reference for very fine anatomical detail.
Through an independent study conducted according to the ISO 5725-1:2023 standard, comparing our models to a reference medical scanner on a panel that includes pathological morphologies.
Yes, it’s an off-the-shelf dataset, already qualified and delivered with its demographic metadata, unlike a custom acquisition which takes several weeks.
