Seraphim Drone Detection Dataset

Open-Sourcing the Future of Counter-Drone AI:

Seraphim Drone Detection Dataset

 
At Seraphim Defence Systems, we believe that the next leap in battlefield protection will be driven by open innovation and AI collaboration.
Today, we’re proud to release the Seraphim Drone Detection Dataset — an open, comprehensive, and one of the largest datasets designed to accelerate research in autonomous counter-FPV systems and AI-based aerial threat detection.

🔗 GitHub: Seraphim Droen Detection Dataset
🤖 Hugging Face: Seraphim Drone Detection Dataset

Why We Built It
FPV and micro-UAV threats evolve faster than traditional countermeasures. Effective defence starts with accurate, fast, and reliable AI vision models capable of spotting small, high-speed drones under complex conditions — urban, rural, day, night, and cluttered environments.

However, the defence AI community has faced the challenge of numerous small, overlapping datasets with duplicated images, rather than a single large, consolidated dataset suitable for this purpose.
Our dataset fills that gap, providing a foundation for training, benchmarking, and improving real-time object detection algorithms integrated into autonomous counter-drone systems like our MACS (Mobile Autonomous Counter-SUAV System).

 
What’s Inside
The Seraphim Drone Detection Dataset includes:

• 83,483 annotated images across multiple drone classes (racing, FPV, quadrotor, and fixed-wing).
• YOLOv8-compatible format, optimized for rapid model training and inference.
• Diverse environmental conditions – from open areas to complex and varied surroundings.
  
• Balanced object sizes, enabling robust detection across drones of varying scales.
• Open licensing for both research and commercial defence applications.

We are continuously expanding the dataset with thermal imagery, low-visibility data, and real-world images from frontline environments to better support multi-sensor fusion research — a critical capability for next-generation counter-UAS systems.

 
Building AI That Protects
The release of this dataset is part of our broader commitment to democratize counter-drone research and accelerate NATO-aligned innovation.
By sharing our data, we aim to empower AI researchers, defence startups, and academic partners to contribute to one of the most urgent challenges on the modern battlefield: detecting and defeating FPV kamikaze drones.

Our vision is simple: AI that saves lives.

 
Join the Mission
We invite developers, engineers, and researchers to:

• Fork and improve the dataset
• Train detection models and share results
• Collaborate on improving precision under real combat conditions

For collaboration or partnership inquiries, contact us.

Follow our journey as we continue to merge AI, autonomy, and real-world defence — from open data to deployable protection systems.

 
“Innovation in defence can’t happen in isolation.
Open data is how we move faster, smarter, and together.”

— Łukasz Grzybowski, Co-Founder & CTO, Seraphim Defence Systems