Find people from the sky.
SkyRescue equips drones with deep learning to spot people in aerial footage, helping rescue teams search large or dangerous areas faster than on foot.
The problem
In search and rescue, every minute counts.
Ground searches over mountains, floods or disaster zones are slow and risky. Drones cover ground quickly, but spotting a person in thousands of aerial frames is hard for human eyes.
precision of our best model (YOLOv8) when detecting humans in drone images.
deep learning detectors benchmarked: RetinaNet, Faster R-CNN, YOLOv5, YOLOv7, YOLOv8.
resolution drone imagery used for training and testing.
Source: the founding research paper (Lecture Notes in Networks and Systems, 2025).
How it works
How it works
A clear path from data to decision.
Fly
A drone surveys the search area and streams 4K imagery.
Detect
A YOLOv8-based model flags likely people in each frame.
Verify
Operators review high-confidence detections on a live map.
Dispatch
Ground teams are sent straight to confirmed locations.
Product
What makes it different.
Every feature is grounded in the founding research.
Benchmarked models
Five detectors compared so the best-performing model leads the system.
High precision
90.12% precision means fewer false alarms for busy rescue teams.
Open toolchain
Built with Python, TensorFlow, Keras and OpenCV.
Continuous improvement
Research shows complex scenes remain hard, guiding our next training rounds.
Who it's for
Who it's for.
Civil defence & rescue teams
Faster coverage of large search areas.
Disaster response agencies
Locate people after floods, earthquakes and fires.
Event & site security
Aerial monitoring of crowds and remote sites.
The research
Built on peer-reviewed research.
- Benchmarks deep learning approaches for detecting humans in drone-acquired aerial images.
- Uses the NTUT 4K drone image dataset.
- YOLOv8 performed best: precision 90.12%, recall 64.34%, F1 75.08%, accuracy 58.75%.
- Complex scenes remain challenging for all models, pointing to future improvements.
Mostafa, S. A., et al. (2025). Benchmarking Deep Learning Approaches for Detecting Humans in Drone-Acquired Aerial Images. Lecture Notes in Networks and Systems, vol. 1293, 239–252. Springer.
Team
The team behind it.
Founded in 2025 by a student team from the faculty below.
Faculty of Engineering (Artificial Intelligence Engineering)
Deep learning, drone integration and field testing.
FAQ
Questions we hear often.
Does the AI replace rescue operators?
No. It highlights likely detections so human operators can verify and act faster.
How accurate is it?
Our best model reached 90.12% precision; research continues to improve recall in complex scenes.
Is this a finished commercial product?
It is a research-based student startup moving from validated prototype toward pilots with real users.
Can we partner or run a pilot?
Yes. We are looking for organisations willing to test the solution in real conditions and share feedback.
Bring SkyRescue to your organisation.
We are onboarding pilot partners now. Read the research and get in touch.