Airbase

The airbase capture.

Six classes across a military airbase, shot from four authored camera zones looking at a target. This is the capture every measured number on this site is taken from, which is the reason it is downloadable: the frames, the instance buffers the labels came from, and the reports that graded it.

Map: Map_Airbase_Demo
Frames in the run: 40
Labelled instances: 4,378
Resolution: 1920×1080
Seed: 67
Validator grade: A · 92.4/100

What it is for

Detection and instance segmentation on cluttered ground scenes seen from the air, and a worked example of what one NameFrame run writes out.

How the labels were made

Every label was derived from the engine's own per-instance ID buffer and the camera transform that rendered the frame — not from a model, and not by hand. Depth is the engine's depth buffer in metres, as float32, not an 8-bit visualisation.

Engine: Unreal Engine 5.8
Camera: Four authored zones looking at a target, 1920×1080, scene weather
Classes: person, car, tank, container, crate, barrel
Annotation formats: YOLO, COCO, per-instance ID buffers, per-frame metadata JSON
Packaged: 13 of 40 frames, across 2 archives
Version: 1.0

What the frames look like

Frames from across the run, downscaled for the web. Nothing here was selected for looking good; they are the report previews the capture wrote.

Aerial frame plugin_000000 of the Unreal Engine airbase capture, with 129 labelled instances outlined by NameFrame — Map_Airbase_Demo1920×1080 per frameReport previews, downscaled for the web

Aerial frame plugin_000003 of the Unreal Engine airbase capture, with 122 labelled instances outlined by NameFrame — Map_Airbase_Demo1920×1080 per frameReport previews, downscaled for the web

Download the archives.

Two cuts of the same run. The frames, labels and metadata are identical where they overlap; what changes is how many frames and how much per frame.

Airbase detection sample

49 MB

A working detection and segmentation set: twelve frames spread across the run, with YOLO and COCO labels, the instance ID buffers they were derived from, and the metadata that says where every camera and object was.

Frames: 12
Annotated instances: 1,813
Format: ZIP

images/ rendered RGB frames, unmodified PNG
segmentation/ per-instance ID buffers, one colour per instance
labels/ YOLO boxes, one text file per frame
annotations/instances.json the same labels in COCO format
metadata/ per-frame camera, actor and spawn-manifest records
reports/ this capture's own report and validation output
capture.json the run contract: classes, camera policy, seed
data.yaml class names, ready for a YOLO trainer

Download 49 MB

SHA-256ac54025db5789bc58e62a54474b3c0ced5563a64c984baf01e155a7e71f17f94

Airbase every-modality sample

14 MB

Fewer frames, everything the engine wrote. Includes the raw float32 depth arrays in metres, which are too large to ship for a whole run but are the thing worth checking if you care about geometry.

Frames: 3
Annotated instances: 562
Format: ZIP

images/ rendered RGB frames, unmodified PNG
segmentation/ per-instance ID buffers, one colour per instance
labels/ YOLO boxes, one text file per frame
annotations/instances.json the same labels in COCO format
metadata/ per-frame camera, actor and spawn-manifest records
reports/ this capture's own report and validation output
capture.json the run contract: classes, camera policy, seed
data.yaml class names, ready for a YOLO trainer
depth/ float32 metres per pixel, as NumPy .npy arrays

Download 14 MB

SHA-256c52922f1ee9db49abd42e00b3ddc25f19dfea83bd4ac0ad01362267d90769f22

Known limitations

The validator graded this run A · 92.4/100, and its report ships inside every archive. Here is what it and we know is wrong with the data.

The verification pass found three visible people with no bounding box, out of ninety it inspected. They lost their pixels to a neighbouring instance in the ID buffer. The failing check ships inside the pack.
Camera placements come from four zones aimed at one target, so viewing angles and distances cover a narrow band. Models trained on this alone will not have seen much variety of viewpoint.
The class balance is uneven, because the scene is uneven. The per-class counts are in reports/report.json.
A sample of 13 packaged frames out of 40 is something to inspect, not a benchmark. Do not report a number against it.

The checks behind those numbers, and how a run is graded, are on the dataset quality page, and the validator itself is described under dataset validation.

Terms, plainly

Free to download, read, train on and benchmark against. Publish whatever results you get, including bad ones. What you may not do is repackage the pack and sell it as your own dataset.

The images depict environment and prop content licensed for use inside Unreal Engine projects. Redistribution rights for that underlying content are not granted with these packs, so check the source licence before publishing derivatives. The full text ships as TERMS.txt inside each archive.

Citation

If you publish anything you got out of this data, this is how to point at the exact run it came from. The seed and the map are the part that matters: they are what makes the run reproducible.

NameFrame. "Airbase: synthetic computer-vision dataset." Version 1.0, generated with Unreal Engine 5.8 on Map_Airbase_Demo, seed 67. https://getnameframe.com/datasets/airbase

Need this scene, but yours?

This capture came out of one NameFrame run: a map, a class list, a camera policy and a seed. Change any of them and you get a different dataset with the same ground truth guarantees. That is what the generator is for.

How synthetic data generation works in Unreal Engine Apply for a private pilot