Real-Time Drone Object Tracking

At Fotokite, I designed a real-time object detection and tracking pipeline for tethered drones used by first responders. The system fuses YOLO-based detection with OSTrack single-object tracking, engineered for motion blur, occlusion, and strict onboard latency budgets.

I benchmarked detection architectures (YOLO, Faster R-CNN, SSD) and built dataset-creation and evaluation pipelines to quantify robustness in real field conditions.

Finally, I optimized inference for embedded deployment, profiling latency/accuracy trade-offs for the drone's onboard compute.