title	CDMS - Crowd Disaster Management System
emoji	🚨
colorFrom	red
colorTo	blue
sdk	docker
pinned	true

Crowd Disaster Management System (CDMS)

An AI-powered crowd monitoring system that detects dangerous crowd density levels in real-time using Computer Vision and Deep Learning.

🏆 Model Performance

• MAE: 2.52 on ShanghaiTech Part B dataset
• RMSE: improved
• Trained custom VGG16-based crowd counting model

✨ Features

• Real-time person detection using YOLOv8
• Custom-trained crowd density estimation model (MAE: 2.52)
• Risk level classification (Safe / Warning / Danger)
• Automatic alert generation and logging
• Live monitoring dashboard

🧠 Pretrained Model

Download the trained model from Hugging Face: 👉 https://huggingface.co/4AK1F/CDMS-crowd-counting

🛠️ Tech Stack

• Python, FastAPI, OpenCV
• PyTorch (custom VGG16-based model)
• YOLOv8 (Ultralytics)
• SQLite, HTML/CSS/JS

🚀 Setup

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Download model from Hugging Face and place in:

model_training/checkpoints/best_model.pth

Run the system:

uvicorn backend.main:app --reload

📊 Dataset

Trained on the ShanghaiTech Crowd Counting Dataset (Part B).

👤 Author

Aakif Mustafa — Final Year CS Project