Developing autonomous systems to neutralize long range drone threats.
My current focus is to train object detection AI model and create real-time object tracking system running on Single Board Computer.
Demonstration of second model I have developed: koshchei-yolo11s-pretrained-v2. It uses same dataset as first model, but cleaned up from mistakes. It's based on default pretrained YOLO11s model to reduce overfitting.
For tracking in the video, I use pure model output without any additional techniques that I plan to introduce later.
Click on the link to video.
Open-Source tools I developed along the way:
-
YOLO Auto Annotation creates YOLO dataset from specified images using Moondream AI model to find bounding boxes for your classes. It reduces time required to go through images and annotate them yourself.
-
YOLO Dataset Visualizer allows to scroll through dataset and see bounding boxes on images. You can remove image from dataset if bounding boxes are incorrect. I use it after Auto Annotation to remove occasional mistakes.
I operate independently. Even though I am student in a university, the institution doesn't provide any resources to me. I share my developments with one of my professors, but that's about it.
My goal is to protect civilian population and infrastructure. Offense systems are discouraged from this project to not compromise moral.
Cool resources
- Coaxial Advantage - shows how coaxial motors/propellers and cylindrical body for drones can be more efficient.