aPrehliadky: offline forensic 3D scanning for iPhone

A photogrammetric 3D scanner for bodies, objects and crime scenes, built for iPhone and iPad. Every frame carries depth and gravity, the scale comes from LiDAR depth and the camera trajectory with an honestly stated uncertainty, and every measurement carries a stated uncertainty: with no upload, no cloud and no account.

Download on the App Store · free Guides and documentation ›

LiDAR + photogrammetry Class A–D certificate T3 target · 247 mm .f3d with SHA-256 Offline · no cloud SK / EN

Why forensic 3D scanning on an iPhone, not a camera and a tape measure

A conventional scene or body examination records a series of 2D photographs with a scale placed in the frame. The limitation is that none of those photographs carries the spatial relationship between findings: the distance between two injuries, the angle of a wound relative to the body's axis, or the position of a body within a room, all have to be reconstructed afterwards from several images and estimation. aPrehliadky removes that step: an iPhone with LiDAR records a depth map and a gravity vector alongside every frame. A photogrammetric model is not in metres by itself: its scale comes from LiDAR depth compared with the model, the camera trajectory (LiDAR + inertial sensors) and from the target triangulated in the model, and every measurement carries an expanded uncertainty U (k = 2).

Reconstruction runs entirely on the device through Apple's RealityKit Object Capture: classical photogrammetry with a LiDAR depth prior, with no NeRF, no 3D Gaussian Splatting and no AI-completed geometry. The report states this explicitly: the model is built exclusively from captured data. For expert practice this is a material distinction from tools that infer missing geometry: aPrehliadky never extrapolates what the camera did not see.

On-model measurement

Straight-line distance, a path along an uneven surface (wound length, circumference), area and volume: in millimetres, each with a stated uncertainty U (k = 2).

Pinned findings

Numbered findings placed directly on the model surface: type (injury, mark, livor mortis, sample), description, movable and editable.

Surroundings from LiDAR

The room mesh as a second, metric, untextured model: the position of a body within a space is a finding in itself.

PDF report and certificate

Case, examiner, capture parameters, a method statement, scale verification, findings, measurements and model integrity in a single document.

How aPrehliadky records an examination

On-scene capture

Two profiles (deceased / living person) and two ranges (whole body / regional detail). The app auto-triggers frames based on camera movement and rotation, and a coverage radar shows what is still missing.

On-device reconstruction

RealityKit Object Capture turns the frame sequence into a textured 3D model: entirely offline, with not a single pixel transmitted.

Scale verification

The T3 target and, when it cannot be laid next to the body, a known dimension in the scene (photo scale, rule, A4 sheet, ISO card) measured in the model as a reference. Without them LiDAR depth provides the scale (class B, about 2 %). The more independent routes, the higher the certificate class.

Review, measurement, findings

Orbit, zoom, a 1:1 AR view, two-tap measurement, and numbered findings pinned directly to the model surface.

Export .f3d and PDF report

A package with a SHA-256 hash of every file for the case record, plus a PDF report carrying the class A–D 3D documentation certificate.

The class A–D certificate and the T3 target

A photogrammetric model is not, by itself, in metres, something else has to supply the scale. aPrehliadky therefore does not award a certificate for any model indiscriminately; instead it rates how many independent routes substantiate that scale: LiDAR depth for the large majority of frames, a fit of the model to the camera's metric ARKit trajectory, and, as a genuinely independent third route, a reference of known length or the T3 calibration target.

The T3 target is a printed pattern on landscape A4 with four circular fiducials and a QR code. The distance between the centres of the opposite fiducials A and B is exactly 247.00 mm: the app locates the fiducials in the model, triangulates their measured distance, and compares it against this reference value. A deviation beyond the allowed tolerance lowers the certificate class or prompts the app to request a fresh printout.

Class A · metrologically verified	LiDAR depth on ≥ 90% of frames, scale corrected by a metrological fit of the trajectory (residual < 25 mm, ≥ 50 cameras) and independent verification by a reference from 290 mm (the T3 target's A–C diagonal, 294 mm, is enough) or the target with error ≤ 1%
Class B · metric	scale from the metrological fit, from LiDAR depth compared with the model (uncertainty about 2%) or LiDAR scale confirmed by a reference or target; independent verification missing or ≤ 2%. Class A always requires a physical verification (T3 target or a reference)
Class C · approximate scale	scale known only approximately (LiDAR without fit correction, or a trajectory coefficient without LiDAR); verification ≤ 5%
Class D · unverified scale	scale not substantiated, or verification above 5%: illustrative documentation only
T3 target	circular fiducials A–B: 247.00 mm · B–C: 160.00 mm · A–C: 294 mm; triangulated in the model, 10 known distances 33–294 mm (A4 print)

First checks in practice

Results of the author's initial tests (not yet a validation study, which is being prepared with a network of institutions):

Reference on a real object: the long side of a label (75 mm) entered as the reference; the label height computed from the model was 35.7 mm, the ruler showed 36 mm. The 0.3 mm difference lies within the expanded uncertainty of ± 5.9 mm (k = 2) reported by the app.
Whole-body simulation manikin: 364 iPhone frames, a detailed model computed on a Mac in about 15 minutes, a captured scene of about 3.6 m. Without a calibration target the app honestly marks the LiDAR scale as indicative (about ± 5 %).
Surface path on a live subject: 12.71 cm ± 0.83 cm across 5 points, with the expanded uncertainty stated for every measurement.

Fotograme3ka 3D: the Mac companion

An older iPhone without LiDAR, or without an A17 Pro chip, can capture in aPrehliadky but cannot compute a model at higher detail on its own. That is what the standalone Fotograme3ka 3D app is for on an Apple Silicon Mac: send the frame package over AirDrop, the Mac recomputes it at medium or full detail, and the model, together with any transferred findings, returns to the iPhone as an .f3d package. More on Fotograme3ka 3D ›

Privacy and data security

No networking layer: the app never connects anywhere and never transmits images or models
Images, models, findings and reports remain on the device until you explicitly export them
The .f3d package carries a SHA-256 hash of every file: the recipient can see on import whether it was modified afterwards
No analytics, no advertising, no account

aPrehliadky is designed as a possible complement to ePrehliadky (the software of the Slovak Health Care Surveillance Authority, ÚDZS): it is an independent application by its author, not a product of ÚDZS. It is not a certified measuring instrument; accuracy must always be verified with a reference of known length or the T3 target.

Frequently asked questions

Is aPrehliadky free? What does the Pro licence add?

Yes. Capture, model computation, measurement, findings, the class A–D certificate and the PDF report are and will remain free, with no limits. The optional Pro licence (a one-time in-app purchase, not a subscription, €149.99) adds only what the free version never had: a cryptographic signature of packages and reports with the device key, the institution name on the report and priority support. Earlier voluntary contributions count towards Pro automatically: from €145 Pro is free, smaller donors get a discounted top-up. Income from Pro and from contributions funds development and research that the author so far finances himself.

Do I need a LiDAR-equipped iPhone for aPrehliadky?

LiDAR is not a strict requirement but is recommended: it gives every frame depth and gravity, the camera trajectory is metric and the resulting certificate reaches a higher class. Without LiDAR, the app derives scale from the ARKit camera trajectory and a calibration target, and the model should be recomputed on a Mac with the Fotograme3ka 3D companion app.

Does aPrehliadky send any data to a server or the cloud?

No. Capture, reconstruction and model review all run entirely on-device through Apple's RealityKit Object Capture. The app has no networking layer: data never leaves the iPhone unless you explicitly export it as an .f3d package.

What is the class A–D 3D documentation certificate?

It is part of the PDF report and rates how many independent routes substantiate the model's scale: LiDAR depth compared with the model, a fit to the camera trajectory, and a reference of known length or the T3 calibration target. Class A denotes the highest degree of metrological verification, class D the lowest, each stated together with its measurement uncertainty.

Is aPrehliadky a certified measuring instrument?

No. Accuracy depends on capture conditions and must be verified with a reference of known length or the T3 target. The app always states this in the report together with the estimated measurement uncertainty.

What is the .f3d package and what is the T3 target for?

The .f3d package bundles the model, findings, measurements and report with a SHA-256 hash of every file, so the recipient can see on import whether the package was modified afterwards. The T3 target is a printed calibration pattern whose fiducial centres sit exactly 247.00 mm apart; the app locates them in the model and uses that distance as an independent scale check.