quailrobot/airspace-v6-sample

QM Airspace v6: Drones, Fixed-wing UAVs & Birds - Free Sample

Buy the full commercial edition: $39 USD -> Polar checkout, instant download
This free sample is non-commercial (CC BY-NC-SA 4.0). The paid full edition has a commercial licence.

Custom dataset of YOUR object ($249)

Need data of YOUR object? Custom Object-Detection Dataset of Your Object (2,000 images), $249 USD -> order on Polar

• What you get: 2,000 labelled photoreal synthetic images (640x640 JPEG) of your own object or scenario (product, part, tool, drone, package, defect...), up to 3 classes, YOLO bounding boxes + data.yaml, train/val/test split, quality report
• Licence: commercial use allowed
• Price: $249 USD one-time; one round of adjustments included
• Delivery: typically 3-5 business days after we receive your reference photos + rough dimensions
• Optional sim-to-real test: send ~200 of your own labelled real images and we report how much the synthetic data improves a detector on them
• Refund: full refund if we cannot deliver your request (14-day refund policy)
• Limits: only objects you own or are allowed to use; no weapons or anything meant to harm people; no copied third-party 3D assets
• Order URL: https://buy.polar.sh/polar_cl_AQu6LzRtqiQmKPlgPt0zePBtvWjJI4vULeFMQ4NLHH9
• Details + contact: https://quailrobot-quailmodel.static.hf.space

Summary

Photoreal synthetic airspace images with YOLO boxes for three classes - multirotor drones, fixed-wing UAVs and birds - over real-world sky and landscape backgrounds, for pre-training drone / counter-UAS detectors that must not confuse drones with birds; the paid full edition can be used commercially.

• Best for: pre-training drone / counter-UAS detectors that must separate drones from birds and small fixed-wing aircraft, before fine-tuning on a small real set
• Not for: deploying a detector trained on synthetic images only; thermal imagery (see Thermal v4)

Quick start

# pip install huggingface_hub ultralytics
from huggingface_hub import snapshot_download
path = snapshot_download("quailrobot/airspace-v6-sample", repo_type="dataset")
from ultralytics import YOLO
YOLO("yolov8n.pt").train(data=f"{path}/data.yaml", epochs=50, imgsz=640)

Facts

• Task: object detection (YOLO format), 3 classes
• Labels: 3 classes: 0 multirotor, 1 fixed_wing, 2 bird; objects less than 20% visible are not labelled; for partly hidden objects the box covers the whole object, including the hidden part; empty label file = true negative (335 of 6,000 images in the full edition, 14 of 300 in the sample)
• Full edition size: 6,000 images 640x640 (train 4,818 / val 603 / test 579) with 14,043 boxes: 6,572 multirotor, 2,494 fixed_wing, 4,977 bird (zip 207.1 MB)
• Free sample size: 300 images (train 246 / val 23 / test 31) with 688 boxes: 320 multirotor, 112 fixed_wing, 256 bird
• Format: 640x640 JPEG images + YOLO txt labels + data.yaml (Ultralytics-ready) + per-object attribute JSON
• What's included (full edition): images/{train,val,test}, labels/{train,val,test}, meta/{train,val,test} (per-object attribute JSON), data.yaml, README.md, LICENSE.txt
• Price: $39 USD; checkout may display the equivalent in your local currency
• Buy URL (primary): https://buy.polar.sh/polar_cl_oraxpjEQru9pU3UPpO3dVaIpwRYv4okp0YXvo0lMPux
• Licence (full edition): LicenseRef-QuailModel-Commercial (commercial use allowed, no resale of the data)
• Licence (free sample): CC-BY-NC-SA-4.0
• Validation: Sim-to-real test (YOLOv8n, 1,000 held-out real drone photos from the Seraphim dataset, CC BY 4.0, evaluation only), multirotor + fixed_wing merged to 'drone', birds unlabelled, mean of 3 training seeds: pre-training on Airspace v6 then fine-tuning on 200 real images = mAP50-95 0.407 (sample SD 0.016 over 3 seeds) vs 0.367 (SD 0.026) for the same 200 real images alone (+4.0 points, +11%; all 3 seeds above all 3 baseline seeds) and mAP50 0.758 vs 0.719. Airspace v6 alone, no real images (1 seed): mAP50 0.420 / mAP50-95 0.206. Caveats: the shot-type mix was tuned using box-size statistics of the real test set (slightly optimistic); the real test set has drones only, so bird and fixed-wing classes are not validated on real photos.
• Data source: Synthetic: original 3D drone, fixed-wing and bird models rendered under real-world CC0 environment lighting from Poly Haven (credited); generator code written with an AI model (AI-assisted)
• Catalog (all QuailModel datasets, catalog.json, llms.txt): https://quailrobot-quailmodel.static.hf.space
• Last updated: 2026-10-10

Validation

Sim-to-real test (YOLOv8n, 1,000 held-out real drone photos from the Seraphim dataset, CC BY 4.0, evaluation only), multirotor + fixed_wing merged to 'drone', birds unlabelled, mean of 3 training seeds: pre-training on Airspace v6 then fine-tuning on 200 real images = mAP50-95 0.407 (sample SD 0.016 over 3 seeds) vs 0.367 (SD 0.026) for the same 200 real images alone (+4.0 points, +11%; all 3 seeds above all 3 baseline seeds) and mAP50 0.758 vs 0.719. Airspace v6 alone, no real images (1 seed): mAP50 0.420 / mAP50-95 0.206. Caveats: the shot-type mix was tuned using box-size statistics of the real test set (slightly optimistic); the real test set has drones only, so bird and fixed-wing classes are not validated on real photos.

Price & licence

• Full edition: $39 USD. One-time payment, instant download after checkout: https://buy.polar.sh/polar_cl_oraxpjEQru9pU3UPpO3dVaIpwRYv4okp0YXvo0lMPux
• QuailModel Commercial Dataset Licence (SPDX: LicenseRef-QuailModel-Commercial): you may train, evaluate and ship models, including in commercial products. You may not resell or redistribute the dataset itself.
• Free sample (this page): CC BY-NC-SA 4.0 - free for non-commercial use.

Limitations

• Synthetic alone does not replace real data (mAP50 0.420 with no real images): pre-train on this, then fine-tune on real images.
• Validation is slightly optimistic: the shot-type mix was tuned on real test-set box-size statistics.
• Bird and fixed-wing classes are not validated on real photos (the real test set contains drones only); fixed-wing models are simple generic airframes.
• Small objects: about 21% of boxes are narrower than 16 px; drone designs are original models, not specific brands.

This free sample: 300 images (train 246 / val 23 / test 31) with 688 boxes: 320 multirotor, 112 fixed_wing, 256 bird. Full commercial edition: 6,000 images 640x640 (train 4,818 / val 603 / test 579) with 14,043 boxes: 6,572 multirotor, 2,494 fixed_wing, 4,977 bird (zip 207.1 MB). Fully synthetic data; summary, facts, validation and limitations are in the block above.

Contents

split	images	boxes
train	246	564
val	23	44
test	31	80

• Image size: 640x640 JPEG. Labels: YOLO txt, classes: 0 = multirotor, 1 = fixed_wing, 2 = bird. Images with an empty label file are true negatives (14 of 300).
• Boxes per class: multirotor 320, fixed_wing 112, bird 256
• meta/ holds per-object attribute JSON (class, box, visibility and module-specific attributes).
• data.yaml included - train directly with Ultralytics YOLO.
• Box size distribution (fraction of image width): median 0.063, 21% of boxes are smaller than 16 px.

How it was made

Original 3D models, procedurally generated and rendered with a physically based renderer under real-world lighting, with realistic camera effects. Labels are computed exactly from the 3D scene (no hand labelling).

Credits

Lighting environments: Poly Haven HDRIs (CC0), credited. Real-photo evaluation set: Seraphim Drone Detection Dataset, (c) 2025 Lukasz Grzybowski, Seraphim Defence Systems (https://huggingface.co/datasets/lgrzybowski/seraphim-drone-detection-dataset), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), used for evaluation only and not redistributed.

Licence

Sample edition: CC BY-NC-SA 4.0 (non-commercial). The full commercial edition is sold by QuailModel (see the buy link).

Disclosure

Generated by QuailModel with AI assistance (generator code written with an AI model); all data is synthetic / computer-generated. Validate on your own real data before production use.