🚀 Deep CargoVision

AI-powered cargo inspection system for automated X-ray analysis, threat detection, and manifest validation using deep learning.

---

🎥 Demo Video

👉 Watch Demo

🧠 Overview

Deep CargoVision is an intelligent cargo inspection platform designed for customs and security operations. It automates the analysis of X-ray cargo images using multiple deep learning models to detect threats, validate manifests, and generate risk insights.

---

🔥 Key Features

• 📦 Single & batch X-ray image analysis
• 🔍 YOLO-based object detection with bounding boxes
• ⚠️ AI-driven risk scoring & anomaly detection
• 📋 Cargo manifest validation (CSV-based)
• 🦎 Wildlife smuggling detection (Reptiles, Birds, Mammals)
• 💧 Liquid container detection (Bottles, Cans, Sprays)
• 📊 Batch analytics & trend visualization
• 🎯 Multi-model inference (YOLO + EfficientNet)
• 🔒 Security-focused dashboard (dark mode UI)

---

🏗️ System Architecture

• Detection Layer → YOLOv8 (object localization)
• Classification Layer → EfficientNet / ResNet18 (refinement)
• Risk Engine → Weighted scoring system
• Validation Layer → Manifest cross-checking

---

⚙️ Tech Stack

Frontend

• React 19 + Vite
• Tailwind CSS
• Framer Motion
• Recharts

Backend

• FastAPI
• Python 3.10+

Machine Learning

• YOLOv8 (Detection)
• EfficientNet / ResNet18 (Classification)
• PyTorch

---

📂 Project Structure

ai-cargo-detection/
├── src/                # Frontend (React)
├── backend/            # FastAPI backend
├── model/              # Trained models
├── training-notebooks/ # Model training
├── public/             # Static files
└── README.md

---

🚀 Quick Start

1. Clone Repository

git clone https://github.com/arpitadhage/Deep_Cargo_Vision.git
cd Deep_Cargo_Vision

2. Frontend Setup

npm install
npm run dev

3. Backend Setup

cd backend
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

---

🧪 Models Used

Model	Purpose
YOLOv8	Object detection
EfficientNet	Image classification
Wildlife Model	Fauna detection
Liquid Model	Container detection

---

📊 Core Capabilities

🔹 Standard Scan (YOLO)

• Fast object detection
• Real-time bounding boxes
• Threat scoring

🔹 Deep Forensic Scan

• Two-stage pipeline (YOLO → Classification)
• Higher accuracy
• Reduced false positives

🔹 Wildlife Detection

• Detects illegal fauna transport

🔹 Liquid Detection

• Identifies hazardous/restricted liquids

---

📡 API Endpoints (Sample)

• POST /predict → YOLO detection
• POST /predict-efficientnet → Deep scan
• POST /predict-wildlife → Wildlife detection
• POST /predict-liquid → Liquid detection
• GET /health → System status

---

📈 Performance Highlights

• Classification Accuracy: ~92%
• Pipeline Accuracy: ~86.4%
• YOLO mAP@50: ~78.5%
• Wildlife Detection mAP: ~96%

---

🔐 Security Notes

• No data stored between sessions
• Local processing supported
• Designed for authorized personnel

---

📌 Future Improvements

• Real-time video stream analysis
• Cloud deployment (AWS/GCP)
• Model optimization for edge devices
• Advanced anomaly detection

---

🤝 Contributing

Contributions are welcome!

• Follow project structure
• Test model changes
• Update documentation

---

⭐ If you like this project

Give it a star ⭐ on GitHub!