♻️ Garbage Image Classifier

AI-Powered Smart Waste Classification System using Deep Learning

A deep learning based application that classifies waste images into different categories and helps promote efficient recycling using Artificial Intelligence.

Built as part of my learning journey during the Edunet Foundation Internship and later upgraded with a production-ready ML pipeline.

---

📌 Overview

Waste management is an important environmental challenge.
This project uses Deep Learning and Computer Vision to automatically classify garbage images into different waste categories.

The system allows users to upload an image of waste material, analyzes it using a trained AI model, and predicts the category along with a confidence score and recycling suggestion.

---

🚀 Live Demo

🔗 Application Link:
https://garbage-classifier-vk18.streamlit.app/

---

📸 Application Screenshots

Homepage

Prediction Result

---

✨ Features

✔️ Upload garbage images
✔️ AI-based waste classification
✔️ Confidence score visualization
✔️ Recycling suggestions
✔️ Fast prediction system
✔️ Clean Streamlit user interface
✔️ REST API support using FastAPI

---

🧠 Classification Categories

The model classifies waste into 6 categories:

Category	Description
📦 Cardboard	Boxes, packaging material
🍾 Glass	Glass bottles/items
🔩 Metal	Cans, metallic waste
📄 Paper	Paper waste
🧴 Plastic	Plastic bottles/items
🗑️ Trash	General waste

---

🤖 Machine Learning Model

The first version of this project used a basic CNN model, but it struggled with real-world images.

The upgraded version uses:

MobileNetV2 Transfer Learning

Architecture:

Input Image
      |
      ↓
Image Augmentation
      |
      ↓
MobileNetV2 Feature Extractor
      |
      ↓
Global Average Pooling
      |
      ↓
Dropout Layer
      |
      ↓
Softmax Classifier
      |
      ↓
Waste Category

---

📊 Model Performance

Final Model Accuracy:

Training Accuracy   : 86%+
Validation Accuracy : 86.34%

Improvements:

• Better generalization
• Real-world image support
• Transfer learning
• Strong image augmentation

---

🛠️ Tech Stack

Machine Learning

• Python
• TensorFlow
• Keras
• MobileNetV2
• Transfer Learning

Backend

• FastAPI
• Uvicorn

Frontend

• Streamlit
• HTML/CSS

Deployment

• Streamlit Cloud

---

📁 Project Structure

Garbage-Classifier-Project/

│
├── backend/
│   ├── main.py
│   ├── predict.py
│   └── requirements.txt
│
├── frontend/
│   ├── app.py
│   └── streamlit_app.py
│
├── model/
│   ├── model.keras
│   └── label_map.json
│
├── notebooks/
│   └── training.ipynb
│
├── utils/
│   └── preprocessing.py
│
├── extra/
│   └── screenshots/
│
├── requirements.txt
├── README.md
└── .gitignore

---

⚙️ Installation & Setup

Clone repository:

git clone https://github.com/deepakjha018/Garbage_Classifier_Project.git

Move into project:

cd Garbage_Classifier_Project

Install dependencies:

pip install -r requirements.txt

Run Streamlit app:

streamlit run frontend/streamlit_app.py

---

🌐 API Usage (Optional)

Start FastAPI server:

cd backend

uvicorn main:app --reload

API endpoint:

POST /predict

Upload an image and receive:

{
    "prediction":"plastic",
    "confidence":95.6
}

---

🔮 Future Improvements

• Add more waste categories
• Improve dataset diversity
• Add object detection support
• Create mobile application
• Deploy FastAPI separately

---

🎓 Internship

This project was initially developed during:

Edunet Foundation Internship Program

Later enhanced with:

• Transfer Learning
• Improved ML pipeline
• Better UI
• Deployment-ready architecture

---

👨‍💻 Developer

Deepak Kumar Jha

AI & Data Science Student

Passionate about Machine Learning, Data Science, and AI-powered solutions.

---

⭐ If you like this project, consider giving it a star!