Military Anthropometric Surveys: Sources and Open Data

This repository collects 49 military anthropometric surveys from around the world. For each survey it records the study report (with a PDF link where one exists), the citation, the corrected survey facts, and the status of the individual-level data. Where that data can be downloaded for free, a copy is included.

• CATALOG.md: the master table of all 49 surveys, grouped by region.
• data/: the four surveys whose individual-level data is freely available, with checksums and notes on units and encoding.
• CORRECTIONS.md: what the September 2026 verification pass changed, and what is still unconfirmed.
• catalog/: the catalog in machine-readable form (surveys.yaml is the source of truth; surveys.csv and surveys.json are generated from it), plus link_status.csv, the latest automated link check.

TL;DR

	Count
Surveys catalogued	49
Individual-level data free to download (copies in data/)	4: ANSUR 1988, ANSUR II 2012, ASRAN 2015, USAF 1967
Free study PDF or open-access paper	22
Commercial data	1 (CAESAR, sold by SAE International)
Entries fully verified / partly verified / not verified	13 / 13 / 23

The four open datasets

Survey	Subjects	Variables	Files	Licence
ANSUR 1988 (US Army)	1,774 men, 2,208 women	131	data/ansur-1988/	Public domain (US Government)
ANSUR II 2012 (US Army)	4,082 men, 1,986 women	93 measured + 14 demographic	data/ansur-ii-2012/	Public domain (US Government)
ASRAN 2015 (Royal Australian Navy)	1,090 men, 232 women	87	data/asran-2015/	CC BY 3.0 AU
USAF 1967 (flying personnel)	2,420 men	202	data/usaf-1967/	US Government data; the R package is GPL (>= 2)

⚠️ Units: In both ANSUR datasets, body mass is stored in hectograms, even in the ANSUR II column named weightkg. Divide by 10 to get kg. Lengths are in mm. The ANSUR II male CSV is Latin-1 encoded. See data/README.md for more.

Quick start

Python

import pandas as pd

ansur2_m = pd.read_csv("data/ansur-ii-2012/ANSUR_II_MALE_Public.csv", encoding="latin-1")
ansur2_m["mass_kg"] = ansur2_m["weightkg"] / 10

ansur88_w = pd.read_csv("data/ansur-1988/ansurWomen.csv", encoding="utf-8-sig")
asran_f = pd.read_csv("data/asran-2015/ASRAN_2015_female.csv")
usaf67 = pd.read_csv("data/usaf-1967/USAF1967.csv")

catalog = pd.read_csv("catalog/surveys.csv")
catalog[catalog.data_access == "public"]

R

ansur2_f <- read.csv("data/ansur-ii-2012/ANSUR_II_FEMALE_Public.csv")
usaf67   <- read.csv("data/usaf-1967/USAF1967.csv", check.names = FALSE)
# The same USAF 1967 data ships with CRAN:
# install.packages("Anthropometry"); data(USAFSurvey, package = "Anthropometry")

Scripts

Install the dependencies with pip install -r requirements.txt.

Command	What it does
python scripts/build_catalog.py	Validates catalog/*.yaml and regenerates surveys.csv, surveys.json, additional_reports.csv and CATALOG.md. Add --check to run it in CI.
python scripts/fetch.py verify	Checks the committed datasets against data/SHA256SUMS. Works offline.
python scripts/fetch.py data	Downloads the datasets again from their original sources, then verifies them.
python scripts/fetch.py reports	Downloads every study PDF in the catalog to reports/<survey-id>/. This folder is git-ignored.
python scripts/fetch.py check-links	Requests every link and writes catalog/link_status.csv, marking each one ok, blocked or broken.
python scripts/build_asran_csv.py	Extracts per-sex CSVs from the ASRAN workbook.
python scripts/build_usaf1967_csv.py	Converts the fixed-width AFFLY67.dat into a CSV with named columns.

About DTIC links. Most study PDFs are hosted on the Defense Technical Information Center (apps.dtic.mil). DTIC and some publishers reject scripted requests from cloud IP addresses with HTTP 403, so the link checker marks their links blocked, not broken. Open them in a browser. DTIC also sometimes shows a "scheduled maintenance" page, so test the links again before you cite them. The study PDFs are linked rather than committed, apart from the report copies that ODL hosts. You can download those with fetch.py reports.

Coverage by region

Region	Entries
United States	14
United Kingdom	4
Canada / Australia / New Zealand	8
Europe / Other	11
Asia-Pacific / Middle East / Latin America	12

Website and database

The website is a static single-page app in site/. It has seven sections:

• Survey catalog. All 49 surveys. You can search them and filter by data access, report access, PDF availability, region, country, service, sexes measured, era, verification and licence, plus year and minimum sample size. Each filter option shows a live count. Results can be sorted, shown as cards or a table, and exported to CSV. DTIC reports carry a link to a free copy on the Internet Archive.
• Explore data. A query builder over the 13,792 people in the four open datasets. You pick surveys, sex, any number of measurement ranges (for example stature 1,700–1,800 mm and BMI ≥ 30), branch and handedness, the columns to show, and how to group results. Results appear as summary statistics, a sortable table of records, distribution charts, and scatter plots with linear fits. Every question is compiled to SQL, which you can view, and can be shared as a link or exported to CSV.
• By country & role. Average measurements of military populations in nearly 60 countries and groupings, from the statistics extracted from about 110 published papers in about 30 languages (aggregates/). A ranking (dot plot and table), a matrix of countries × measures, and a per-population view where every row links to the paper and page it came from. Filter by measure, sex, country, role and minimum sample size.
• SQL console. Free-form SQLite queries against the whole database, with a schema browser, example queries, sortable results and CSV export.
• Measures. The 59 harmonized measures, showing which surveys include each one and the mean for men and women.
• Papers. 258 papers and reports in 30 languages (foreign-language titles are shown with an English translation), with free links and the amount of data extracted from each.
• About. Where the data comes from, how it was harmonized, caveats, and downloads.

The database is described in database/README.md. The browser downloads it once (about 5 MB compressed) and runs every query locally with sql.js, so no server is needed.

python scripts/build_database.py        # database/anthro.sqlite only
python scripts/build_site.py            # database + site into public/
python -m http.server -d public 8000    # preview at http://localhost:8000

Papers and Google Drive. python scripts/fetch.py reports downloads every paper into reports/ (git-ignored, about 600 MB; DTIC blocks scripts, so it falls back to the Internet Archive copy). python scripts/build_papers.py writes the index catalog/papers.csv. A Google Drive folder holds the paper index and a country/role summary as Google Sheets (private, shared on request). Drive cannot import files from a URL, so scripts/drive_import.gs is a Google Apps Script that copies all the open PDFs into that folder from your own account; see the README in the folder.

To deploy on Vercel: choose Add New → Project, import this repository, and keep the settings from vercel.json (no framework, build command python3 scripts/build_site.py, output directory public). Every push to main redeploys the site.

Digital human (3D)

The Digital human tab (site/js/human/, design in docs/human3d/DESIGN.md) builds a posable 3D mannequin from a population, a real person, an imported file or a manual stature and mass: a 77-joint rig with ranges of motion, a skin generated from the measured girths, the mapping of every measure to every bone and a fit report (MODEL.md). It has a library of 200+ postures (POSE.md), a joint and IK posture editor (EDITOR.md, IK.md), a sandbox with seats, displays, boxes and tools, sit/grab/reach placement and distance measurements (PLACE_ANALYSIS.md), posture angle analysis (METRICS.md) and export to GLB, OBJ, STL, BVH, CSV and an exchange JSON (EXPORT.md). Every posture, scene object and interaction (sitting, a hand on a grip, looking at a display, lifting a load) is tagged MIL-STD-1472H compliant, not compliant or to check, with the clauses that decide it (). three.js r180 (MIT) is vendored in . Tests: (Node 22; browser tests: ).

MIL-STD-1472H requirements

standards/mil-std-1472h/ lists 1,726 requirements of MIL-STD-1472H (DoD Human Engineering, 15 September 2020) that the website could check or report. Each requirement gives its paragraph, page, numeric limits and how the Digital human tab would check it. That covers stairs, ladders, doors, hatches, consoles, seats, displays, labels, reach, range of motion, lifting, carrying, handles, controls, valves and vehicle stations. The README explains the standard's accommodation rules, maps the requirements to the site's features, lists where the current sandbox ranges disagree with 1472H, and proposes a build order. python scripts/build_standards.py validates the list and regenerates the CSV.

API and MCP server for CAD and digital-human tools

anthro_service/ serves the country × role × sex percentiles (P5/P50/P95 or any percentile of up to 571 measures, with reliability and sources) to tools that build mannequins and digital humans (CATIA Human Builder, Jack, RAMSIS, AutoCAD scripts, Blender) and to AI assistants:

• Static files at /api/v1/… on the deployed site (python scripts/build_site.py writes them): JSON and CSV per population, no server.

• REST API with an OpenAPI description, CSV and XML output: pip install -r anthro_service/requirements.txt && python -m anthro_service serve.

• MCP server for Claude Desktop, Claude Code, Cursor and other MCP clients: python -m anthro_service mcp.

• Send to CAD: the Digital human's Export tab pushes the figure you built (measures, dimensions, posture) to this server (POST /v1/profiles); CATIA and other tools read it back from /v1/profiles/latest/values, /angles and /points (or the VBA functions AnthroProfile, AnthroProfileAngles), and AI assistants through the MCP tools list_pushed_profiles and get_pushed_profile. Without a browser: python -m anthro_service push FILE.json.

See anthro_service/README.md for the endpoints, the percentile and reliability definitions, examples (Python, VBA, PowerShell, curl) and the licence caveat for commercial use.

Contributing

To add a source or fix an entry:

1. Edit catalog/surveys.yaml.
2. Run python scripts/build_catalog.py.
3. Commit the YAML together with the regenerated files.

If you add a dataset, put it under data/<survey-id>/, add its local: path to the catalog entry, and update data/SHA256SUMS.

Licences

The datasets keep their original licences, listed above and in data/README.md. If you use ASRAN, you must credit the Commonwealth of Australia (Defence Science and Technology Group) under CC BY 3.0 AU. The US Government reports and datasets are in the public domain. The NZDFAS monograph is CC BY-NC 4.0.