🚮 LitterVision

---

🌍 Live Demo

🔗 https://litter-vision.onrender.com

⚠️ Free hosting may take ~30 seconds to wake up (cold start)

---

📌 About the Project

LitterVision is an AI-powered litter detection and classification system that uses computer vision and deep learning to identify different types of waste from real-world images.

The system helps quantify litter severity and provides environmental guidance, making it suitable for urban cleanliness and smart city initiatives.

---

🧠 Problem Statement

Improper waste disposal is a major contributor to urban pollution. Manual monitoring of litter is inefficient and costly.

Solution:
An automated computer vision system that can:

• Detect litter presence
• Classify waste type
• Quantify cleanliness level
• Assist urban cleanliness decision-making

---

⚙️ Tech Stack

Layer	Technology
Frontend	HTML, CSS (Glassmorphism UI)
Backend	Flask
ML Model	MobileNetV2 (Transfer Learning)
Framework	TensorFlow / Keras
Deployment	Render (Gunicorn)

---

🧪 Features

✅ Image-based litter classification
✅ Camera capture support (mobile-friendly)
✅ Confidence score with animated bar
✅ Cleanliness severity quantification
✅ Environmental tips based on litter type
✅ Modern animated UI
✅ Deployed with public URL

---

🗂️ Litter Categories

• 🟫 Cardboard
• 🟦 Glass
• 🔩 Metal
• 📄 Paper
• 🧴 Plastic
• 🗑️ Trash

---

📊 Cleanliness Quantification Logic

Confidence Score	Cleanliness Level
0–30%	🟢 Clean Area
31–70%	🟡 Moderately Polluted
71–100%	🔴 Highly Polluted

This satisfies litter quantification for urban cleanliness analysis.

---

📸 Application Flow

User uploads image / uses camera
        ↓
Image preprocessing
        ↓
Deep Learning Model (MobileNetV2)
        ↓
Prediction + Confidence
        ↓
Cleanliness Score + Environmental Tip

---

🚀 Deployment

The application is deployed on Render (Free Tier) using:

gunicorn app:app --workers 1 --threads 1 --timeout 120

Optimized for low-memory cloud environments using:

• CPU-only inference
• Lazy model loading

---

🧠 Viva / Interview Explanation

“LitterVision is a computer vision–based system that classifies litter from real-world images and quantifies cleanliness levels using confidence-based severity scoring, supporting urban cleanliness initiatives.”

---

📁 Project Structure

LitterVision/
│
├── app.py
├── train.py
├── model.h5
├── requirements.txt
├── templates/
│   └── index.html
├── static/
│   └── uploads/
│       └── favicon_io/
├── Dataset/
│   ├── cardboard/
│   ├── glass/
│   ├── metal/
│   ├── paper/
│   ├── plastic/
│   └── trash/
└── README.md

---

🛠️ How to Run Locally

pip install -r requirements.txt
python app.py

Open:

http://127.0.0.1:5000

---

🌱 Future Enhancements

• Bounding-box litter detection (YOLO)
• Prediction history dashboard
• Area-wise cleanliness analytics
• Grad-CAM explainability
• Mobile app integration

---

👨‍💻 Author

Vansh Agrawal Engineering Student | AI & ML Enthusiast

🔗 GitHub: https://github.com/vansh070605

⭐ If you like this project, consider starring the repo!