Python YOLOv8 OpenCV Streamlit Plotly W&B License

Production-grade object detection and tracking - YOLOv8 + BoT-SORT/ByteTrack, interactive Streamlit dashboard, Plotly analytics, MP4 export and optional W&B logging.

Built by Md. Musa Islam Fahad · CSE (Data Science) · Daffodil International University

📖 Overview

✨ Features

🧰 Tech Stack

📁 Project Structure

object_detection_tracker/
│
├── app.py # Streamlit main application — entry point
│
├── src/
│ └── tracker.py # ObjectTracker class (YOLOv8 + BoT-SORT engine)
│
├── utils/
│ ├── analytics.py # SessionStats + Plotly chart builders
│ ├── video_utils.py # Video I/O helpers (read, write, frame extraction)
│ └── logger.py # W&B experiment logger
│
├── .streamlit/
│ └── config.toml # Streamlit theme + server configuration
│
├── requirements.txt # Python dependencies
├── Dockerfile # Container build definition
├── .env.example # Environment variable template
├── .gitignore
└── README.md

⚙️ Local Installation

1. Clone the repository

git clone https://github.com/MusaIslamFahad/codealpha_tasks.git
cd codealpha_tasks/CodeAlpha_Object_Detection_and_Tracking

2. Create a virtual environment

python -m venv venv
# On Windows
venv\Scripts\activate
# On macOS / Linux
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

pip install ultralytics opencv-python streamlit plotly wandb python-dotenv

4. Configure environment variables (optional - for W&B)

cp .env.example .env
# Edit .env and add: WANDB_API_KEY=your_key_here

5. Run the app

streamlit run app.py
# Open http://localhost:8501

🚀 Usage

Streamlit Dashboard

streamlit run app.py

Keyboard Controls (webcam mode)

🐳 Docker Deployment

# Build the image
docker build -t visiontrack-ai .
# Run on CPU
docker run -p 8501:8501 visiontrack-ai
# Run with GPU
docker run --gpus all -p 8501:8501 visiontrack-ai
# Open http://localhost:8501

☁️ Deploy to HuggingFace Spaces

git remote add hf https://huggingface.co/spaces/YOUR_USERNAME/visiontrack-ai
git push hf main

☁️ Deploy to Streamlit Cloud

📊 W&B Experiment Tracking

🧠 Architecture

Input (Image / Video File / Webcam)
 │
 ▼
 ┌─────────────┐
 │ YOLOv8 │ ← Pre-trained on COCO (80 classes)
 │ Detector │ ← Configurable: n / s / m / l / x variant
 └──────┬──────┘
 │ raw detections (bbox, class, confidence)
 ▼
 ┌──────────────────────────┐
 │ Tracker │ ← BoT-SORT: appearance embedding + Kalman filter
 │ (selectable from UI) │ ← ByteTrack: motion-only, lighter & faster
 └──────┬───────────────────┘
 │ tracked detections (bbox, class, track_id)
 ▼
 ┌─────────────────────┐
 │ Annotation Engine │ ← Bounding boxes, colour-coded IDs, trails, FPS HUD
 └──────┬──────────────┘
 │
 ▼
 ┌───────────────────────────────┐ ┌────────────────┐
 │ Streamlit Dashboard (app.py) │────▶│ W&B Logger │ (optional)
 │ ├─ Live annotated feed │ │ (logger.py) │
 │ ├─ Plotly analytics charts │ └────────────────┘
 │ │ ├─ FPS timeline │
 │ │ ├─ Object count chart │
 │ │ ├─ Class breakdown │
 │ │ └─ Track timeline │
 │ └─ MP4 download button │
 └───────────────────────────────┘

🎯 Model Variants

# Swap model by changing one line in app.py or src/tracker.py
model = YOLO("yolov8n.pt") # Nano - fastest, lowest VRAM
model = YOLO("yolov8s.pt") # Small
model = YOLO("yolov8m.pt") # Medium
model = YOLO("yolov8l.pt") # Large
model = YOLO("yolov8x.pt") # Extra-large - most accurate

🔁 Tracker Comparison

📋 Requirements

ultralytics>=8.0.0
opencv-python>=4.8.0
torch>=2.0.0
streamlit>=1.35.0
plotly>=5.18.0
wandb>=0.16.0
python-dotenv>=1.0.0
numpy>=1.24.0

📦 COCO Classes (80)

🤝 Contributing

👤 Author

📄 License

🙏 Acknowledgements

⭐ If you found this useful or built something cool on top of it, drop a star. It helps a lot!

⬆ Back to Top