🎯 Object Tracking with OpenCV DNN

---

Track objects in real-time using multiple OpenCV legacy trackers.

This project demonstrates object tracking through a layered approach:

• A basic single-tracker demo
• Tracker-specific test modules
• Side-by-side multi-tracker comparison

---

✨ Key Features

• 📦 Modular tracker scripts (CSRT, KCF, BOOSTING, etc.)
• 🎥 Video output with bounding boxes & FPS overlays
• 📊 Visual comparison of tracker performance
• ⚙️ Easy to extend with new videos or trackers

---

🏡 Folder Structure

object-tracking-opencv/
│ 
├── assets/
│   ├── banner.png
│   └── sample_output.png
│
├── main/                          # Initial Demo & Tracker Comparison
│   ├── intro_demo.py
│   └── comparison.py
│ 
├── trackers/                      # Different types of trackers
│   ├── boosting_tracker.py
│   ├── csrt_tracker.py
│   ├── kcf_tracker.py
│   ├── medianflow_tracker.py
│   ├── mil_tracker.py
│   ├── mosse_tracker.py
│   └── tld_tracker.py
│ 
├── utils/
│   └── helper_functions.py
│ 
├── input/                        ← place input .mp4 videos here
│   └── race_car.mp4
│ 
├── output/                       ← auto-saves results
│   ├── comparison_output.mp4     (More demo videos are available inside this folder)        
│   └── kcf_output.mp4
│ 
├── requirements.txt
├── .gitignore
├── LICENSE
└── README.md

---

🚀 How It Works

🔹 Single Tracker Demo

python main/intro_demo.py

🔹 Tracker-Specific Run

python trackers/csrt_tracker.py

🔹 Multi-Tracker Comparison

python main/comparison.py

Results will be saved inside the output/ folder as .mp4 videos.

---

📋 Tracker Modules Available

• BOOSTING
• CSRT
• KCF
• MEDIANFLOW
• MIL
• MOSSE
• TLD

All use OpenCV’s cv2.legacy API and run in real-time.

---

🔢 Tracker Comparison Table

Tracker	🔁 Speed	🎯 Accuracy	⚖️ Robustness	📌 Best For	Notes
BOOSTING	🟠 Slow	🔴 Low	🔴 Weak	Basic demos, low-demand tasks	Based on AdaBoost; outdated
MIL	🟡 Medium	🟡 Medium	🟡 Okay	Less jittery than BOOSTING	Handles occlusion better than BOOSTING
KCF	🟢 Fast	🟢 Good	🟠 Not for scale changes	High-speed, consistent tracking	Efficient but fails on scale/rotation
TLD	🔴 Very Slow	🟡 Medium	🟠 Fragile	Academic experiments	Learns online, prone to drift
MEDIANFLOW	🟡 Medium	🟢 High	🔴 Weak on occlusion	Short sequences with smooth motion	Fails on fast motion
MOSSE	🟢⚡ Fastest	🟠 Lower	🟠 Fragile	Real-time apps, fast motion	Great FPS, low robustness
CSRT	🔴 Slowest	🟢✅ Best	🟢 Strong	High accuracy tasks, low-FPS OK	Most accurate, but slowest

📌 Tracker comparison is based on documented benchmarks, developer experience, and OpenCV literature

---

📦 Installation

pip install -r requirements.txt

---

🖼 Output Examples

“From bounding boxes to beauty — watch your object come alive through tracking.”

---

Sample Demos (click to download)

• KCF Demo
• Comparison Demo

👉 More demo videos are available in the output/ folder.

---

📄 License

MIT License © 2025 Heleena Robert.
You’re free to use, modify, and share with ❤️

---

👩‍💻 Author

Heleena Robert
GitHub