Why does the quality of your 3D database determine the success of your innovation? Whether you’re developing a custom medical device, a retail fitting application, or a morphological AI algorithm, raw data is no longer enough. To be usable in R&D, a 3D dataset has to meet strict metrology requirements. Here are the three criteria to check to know how to choose a 3D dataset — or buy one.
In short
A good 3D dataset is judged on three criteria: the geometric precision of the mesh (integrity, topological correspondence between models), the repeatability of the acquisition protocol (standardized pose, automated cleanup), and the richness of the associated metadata (age, sex, weight, height, ethnicity). MyFit Solutions validates its precision according to the ISO 5725-1:2023 standard, with a scale accuracy of around 99%.
3D mesh precision: far more than a polygon count
Mesh precision is often confused with resolution. Yet for an R&D engineer, a “heavy” mesh with several million polygons isn’t necessarily precise.
Geometric integrity comes first
3D mesh quality starts with geometric integrity: a watertight mesh dataset is a must for volume calculations and physical simulations. At MyFit Solutions, our reconstructions are delivered as triangulated meshes, the de facto standard for 3D scanning pipelines (photogrammetry, lidar, structured scanners). What really makes the difference for statistical analysis isn’t the type of facet, but mesh topology 3D dataset correspondence between models: a good dataset guarantees that every scan shares the same number of vertices and the same connectivity, through registration onto a common template. Without this point-by-point correspondence, comparing individuals with one another is impossible.
Signal-to-noise ratio: precision that can be proven
Precision is measured by the absence of sensor-related artifacts. A mesh accurate to within a few millimeters, depending on the area being analyzed, matches the standard expected for health and ergonomics applications.
Our reconstruction engine has been validated according to the ISO 5725-1:2023 standard: an average coefficient of variation of 1.1% for repeatability, an average relative deviation of 1.1% compared to a reference medical scanner, for a scale accuracy of around 99%, including on complex pathological morphologies.
Repeatability: the key to a usable 3D database
A large 3D database is an asset, but it’s the consistency of the data that gives it its value. Repeatability guarantees that every scan was captured using the same protocol, a necessary condition for comparing individuals without methodological bias.
- Pose standardization: a uniform orientation in space, to compare thousands of individuals without morphological bias.
- Automated cleanup: on large volumes, automatically filtering outliers and filling holes, to keep a consistent dataset.
Metadata: what turns a dataset into an analysis tool
A high-performing 3D dataset is enriched data. Geometry alone is no longer enough once you want to segment a population: metadata becomes the real analytical lever.
- Sex and age: to track morphological change across populations.
- Weight and height: to correlate 3D shape with health indicators such as BMI.
- Segmentation: a well-structured database makes it possible to isolate specific segments to refine product design.
In short, a good 3D dataset is judged on three criteria: the geometric precision of the mesh, the repeatability of the acquisition protocol, and the richness of the associated metadata.
Discover our data offers
The 3 criteria at a glance
| Criteria | What to check | Why it matters |
| 3D mesh precision | Manifold (watertight) mesh, topological correspondence between models | Determines volume calculations, simulations, and comparison between individuals |
| Repeatability | Standardized acquisition protocol, uniform pose, automated cleanup | Ensures bias-free comparison between scans |
| Metadata | Age, sex, weight, height, ethnicity, geographic region | Turns raw geometry into a segmentation and analysis tool |
Summary table of the 3 criteria for choosing a 3D dataset.
To understand how this data is acquired in the field, our article on our 3D data acquisition methods walks through the protocol.
Our solutions to accelerate your R&D
Building a proprietary 3D database is a long and costly process. MyFit Solutions simplifies access to high-precision anthropometric data, in two complementary ways.
Access our exclusive 3D datasets
Save months of research with datasets that are already qualified, cleaned, and segmented, available as raw or meshed data:
- Heads (tight cap): 800 scans, with complete demographic metadata (Europe / Middle East).
- Heads (multi-ethnic): 100 scans with balanced ethnic quotas (Caucasian, Middle Eastern, African) and more than 150 images per model.
- Ears: 1,000 scans, spread across North America, Europe, and Asia-Pacific, delivered with anatomical landmarks.
- Hand-wrist: 1,000 scans with balanced ethnic quotas, for orthopedic fitting or wearables studies.
- Legs (pairs): 100 scans with balanced ethnic quotas and more than 150 images per model.
- Open-fingered hands: 100 scans, Caucasian morphology, with more than 100 images per model.
- International sourcing: need a specific population segment? Our global partner network makes it possible to acquire a custom dataset.
Each dataset is delivered with its associated metadata (at minimum: ethnicity, geographic region, gender, age), and technical characteristics such as volume are adapted to each project.
Build your own database
Prefer to build your own dataset? Our software platform supports companies in health and every other sector at each step: capture via mobile or professional scanners, smart manual measurement tools for expert validation, and automated measurement at scale for your R&D statistics.
MyFit Solutions carries out more than 100,000 3D scans and digital measurements every year, used daily in more than 15 countries, in the health (medical devices, MedTech) and retail sectors.
Frequently asked questions
MyFit Solutions offers both options. We provide a software solution that lets you build your own standardized databases, usable for R&D. Failing that, we collect data all over the world, across different populations, for your machine learning and AI analysis projects, or simply to build a morphological measurement database.
For datasets that are already available (heads, ears, hands…), delivery is nearly immediate. For a custom need, such as a specific population or a particular volume, our team and our partner network can deliver several thousand 3D models within a few months. The exact timeline depends on your technical and metadata requirements.
Depending on your needs, we deliver raw or already-meshed data, together with the associated metadata (at minimum: ethnicity, geographic region, gender, age). Post-processing options, such as cropping, smoothing, and landmarking, can be added on request.
