The Challenge
Detect small, fast-moving drones in varied outdoor scenes with reliable bounding boxes and low-latency overlays for live camera feeds.
The Solution
We built a Flask web app that serves a live MJPEG camera stream and processes uploaded images/videos through a YOLOv11 model (`models/best.pt`). The app draws bounding boxes, exposes a `/video_feed` stream, and provides downloadable processed outputs.
Impact
The prototype achieves accurate drone localization on the provided Roboflow dataset and includes training notebooks so the model can be refined for production.
