New Delhi (ABC Live): YOLOv10 for Real-Time Military Aircraft Detection: Military aircraft detection in aerial imagery is now a core need for defence. Spotting fighter jets, bombers, and drones quickly helps nations guard their skies, improve watch systems, and respond faster to threats.

Unlike civilian flight tracking, military aircraft detection must work in tough settings—bright sun, clouds, or tilted angles. This is why testing new AI systems is vital. ABC Live Research looks at how YOLOv10 improves this task, and how it compares with earlier versions and global research.

Dataset and Method

• Dataset: 74 types of aircraft (F-35, B-52, Rafale, Su-57) with box labels.

• Conditions: Day, cloud, and other real-world cases.

• Models Tested: YOLOv5, YOLOv8, and YOLOv10.

• Measures: Precision, Recall, F1-score, mAP, and speed.

Results for Military Aircraft Detection in Aerial Imagery

YOLOv10 Results

YOLOv10 reached a Precision of 82%, a Recall of 66.4%, and an F1-score of 0.687. It also gave an mAP@0.5 of 76.4% and mAP@0.5–0.95 of 68.7%. Most importantly, it worked in real time with only 3.8 ms per image.

Comparison

As a result, YOLOv10 proves both faster and more accurate than its earlier versions.

Other Research on Aircraft Detection

Most studies focus only on scores. However, they use different datasets, such as:

• RarePlanes: Real + made-up images to fill data gaps.

• DOTA-v2: Rotated boxes for angled objects.

• DIOR: Boxed dataset with airplanes.

• xView: Many aircraft classes, large scale.

• FAIR1M: Fine aircraft sub-types.

• UCAS-AOD / NWPU VHR-10: Early but still used benchmarks.

Therefore, ABC Live adds value by linking these results with defence and policy, not just numbers.

Diagnostics

• Error Maps: Best on bombers and stealth jets, weaker on look-alike fighters.

• PR Curves: Good at precision but less recall with hidden aircraft.

• Also, box marks stayed accurate even in low-contrast scenes.

Strategic Impacts

1. Defence Work: Speeds up early threat alerts.

2. Airspace Safety: Key for border and conflict zones.

3. Automation: Cuts down on human image checks.

4. Thus, faster choices are possible for commanders.

Limits and Next Steps

• Hard to tell apart aircraft that look similar.

• Although strong in daylight, weaker at night or in heavy clouds.

• Adding radar or infrared could help.

• Also, the ethics of AI in war need debate.

Why This Report is Different

• Unlike other studies, it compares across datasets and models.

• Also, it ties AI gains to defence and public policy.

• While most work centres on the U.S. or China, ABC Live looks at India and the Global South.

• In short, it mixes tech depth with clear language for all readers.

Editorial Note

At ABC Live, we aim to give more than data. We turn research into insights that guide policy, defence, and democracy. This report on military aircraft detection in aerial imagery shows how YOLOv10 works, places it in global context, and links it to India’s real needs.

Thus, ABC Live stands as a bridge between technology and public debate.

— Jatinder Kaur 
Editor, ABC Live

Sources

1. RarePlanes Dataset – https://arxiv.org/abs/2006.02963

2. DOTA-v2 – https://captain-whu.github.io/DOTA/

3. DIOR Dataset – https://github.com/Gaozhongpai/DIOR

4. xView Dataset – https://xviewdataset.org/

5. FAIR1M Dataset – https://arxiv.org/abs/2103.05569

6. UCAS-AOD – https://github.com/ucas-vg/UCAS-AOD

7. NWPU VHR-10 – http://www.escience.cn/people/JunweiHan/NWPUVHR-10.html

8. YOLOv10 GitHub – https://github.com/THU-MIG/yolov10

9. Object Detection Survey – https://arxiv.org/abs/2106.13230