Skip to content
NAMEFRAMECommercial PreviewApply for Pilot
Quality

Every number here came from one real run.

This page uses a single 40-frame capture of Map_Airbase_Demo, taken on Unreal Engine 5.8 and processed by the same pipeline a customer would use. The distributions are recomputed from its buffers and manifests; the checks are quoted from its report. Nothing was selected to look good.

Frames
40
Instances
4,378
Dataset grade
A · 92.4/100
Raw health
B · 82/100
Seed
67
Quality

More data is not the same as better data.

Ten thousand frames of the same view from the same distance is one frame, repeated. These are the real distributions of the 40-frame airbase capture this page has been showing, computed from its own buffers and manifests.

Measured
4,402
Min
33.19m
Median
78.87m
90th
108.06m
Max
187.81m

Every annotated instance, measured from the camera position to the actor's own recorded position. The run's cameras sit in four authored zones, which is why the mass falls between 40 and 120 metres rather than spreading evenly.

Target, actual, gap

Only constraints this run actually declared appear here. A target nobody configured would be a number invented for a table.

ConstraintTargetActualGap
Minimum spacing between placements≥ 1.00 m1.00 mmet
Viewpoint retries per frame≤ 329met
Targets visible per accepted frame≥ 1873met
Frames with no annotation00met

Every check, pass or fail.

Quality gates decide whether a capture is allowed to become a dataset. Verification checks go further and ask whether the labels agree with the geometry that produced them. Both run on every capture, and both are recorded whether or not the answer is flattering.

12/13checks passed on this run. The one that did not is listed below with the rest.
CheckKindResultRecorded message
empty-frame-rateQuality gatePASS0/40 empty frames (0.0%, max 0.0%)
empty-seg-rateQuality gatePASS0/40 empty masks (0.0%, max 0.0%)
total-instancesQuality gatePASS4378 instances (min 1)
mean-instances-per-frameQuality gatePASS109.45 instances/frame (min 0.10)
black-rgbQuality gatePASS0 black RGB frames (max 0)
missing-filesQuality gatePASS0 missing files (max 0)
small-object-rateQuality gatePASS40/40 frames with boxes under 12px (100.0%, max 100.0%)
determinismVerificationPASS2 frames labelled twice, byte-identical
configuration-snapshot-integrityVerificationPASSconfiguration snapshot verified (e940fe63a1198cb9abc634330de558b19de566e60ec4c6fa0043b71001110f9e)
spawn-manifest-integrityVerificationPASS40 frame manifests and index agree; 3360 candidates, 0 rejected with reasons
color-collisionsVerificationPASS289 colours across 289 classes, no collisions
pixels-vs-transformsVerificationPASS197 blobs match their recorded actor positions
missing-visible-peopleVerificationFAIL3/90 visible people have NO box (3.3% false negatives, likely cross-paint victims): 41, 35, 24

The failing check found three visible people with no bounding box, out of ninety in the sample it inspected. They lost their pixels to a neighbouring instance in the ID buffer. A dataset that hid this would still contain the error; the only difference would be that you found it later.

Output

Unreal goes in. A dataset comes out.

The engine writes a raw dump and nothing else. Everything a trainer consumes is derived from that dump afterwards, which is why the same run can be re-exported into a different format without going back to the editor.

SourceUNREAL

Map_Airbase_Demo · 40 frames · 1920×1080

On disk
  • raw/rgb.png, seg.png, depth.npy per frame
  • labels/boxes and masks derived from the ID buffer
  • metadata/frame.json, spawn_manifest.json, capture.json
  • reports/report.json, validation.json, previews
  • dataset.yamlclass names and split paths
Exported
COCOYOLORAW
SplitImagesLabelsMasks
train313131
val999
test000

Only the formats this build genuinely writes are listed. The test split is empty on this run because it was exported with a train and validation split only, and an empty split is shown as empty rather than quietly dropped.

Or skip the argument and read the files.

The frames, masks, labels, metadata and the reports quoted on this page are downloadable. Open them, run your own checks, and decide for yourself.

Download the sample datasets →