Drone-Based Object Detection and Tracking (Simulated)

A real-time object detection and tracking system for aerial surveillance using YOLOv11, designed for detecting and tracking vehicles (civilian/military), aircraft, and maritime vessels from drone footage.

Overview

This system implements state-of-the-art object detection using the YOLOv11 model, trained on aerial imagery datasets, achieving high-accuracy real-time detection across 6 object classes with comprehensive tracking capabilities.

Key Features

• Real-time object detection at 30+ FPS
• Multi-object tracking using ByteTrack algorithm
• 6-class detection: Cars, Buses, Trucks, Boats, Aircraft, Military Vehicles
• Comprehensive evaluation framework with per-class metrics
• Production-ready inference pipeline with visualization dashboard (not yet implemented)

Tech Stack

• Framework: Ultralytics YOLOv11
• Model: YOLOv11-Small (optimized for accuracy/speed balance)
• Training Platform: Google Colab (T4 GPU)
• Languages: Python 3.10+
• Key Libraries: PyTorch, OpenCV, NumPy

Dataset

• VisDrone2019: 7,000+ drone-captured images
• Custom Dataset: 134 manually annotated images (via DroneStock, annotated with CVAT)
• Total: ~7,200 images

Training Platform

• Hardware: Google Colab T4 GPU (16GB VRAM)
• Duration: 2 hours
• Framework: Ultralytics 8.3.x

RESULTS

(go to training_report.md in docs directory for more detailed analysis)

Overall Performance

Metric | Score
| mAP@50 | 80.3% |
| mAP@50-95 | 62.4% |
| Precision | 88.0% |
| Recall | 75.7% |

Per-class Performance

Car: 84.0% mAP@50
Bus: 60.5% mAP@50
Truck: 38.9% mAP@50
Boat: 99.5% mAP@50
Aircraft: 99.5% mAP@50
Military: 99.5% mAP@50

Dataset Citation

@ARTICLE{9573394, author={Zhu, Pengfei and Wen, Longyin and Du, Dawei and Bian, Xiao and Fan, Heng and Hu, Qinghua and Ling, Haibin}, journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, title={Detection and Tracking Meet Drones Challenge}, year={2021}, volume={}, number={}, pages={1-1}, doi={10.1109/TPAMI.2021.3119563}}