📦 Smart Waste Classification

1. Screenshots

1.1 The App Interface

The Gradio-based web application for waste classification.

1.2 Classification Result

The classification output showing the predicted waste category.

---

2. Overview

This project is an AI-powered trash classification system that identifies different types of waste: cardboard, glass, metal, paper, plastic, and trash. It includes:

• Dataset preparation & upload to Hugging Face.
• Model training using ResNet18.
• Deployment of a Gradio app on Hugging Face Spaces.

🔗 Live Demo: Smart-Waste-Classifier-App

---

3. Table of Contents

1. Key Features
2. Technology Stack
3. Project Structure
4. Installing & Running
5. Processing Pipeline
6. Data Analysis
7. Deployment Steps
8. References
9. License

---

4. Key Features

• Automated Waste Classification

  • Supports cardboard, glass, metal, paper, plastic, trash.
  • Uses ResNet18 for classification.

• Robust Data Processing

  • Dataset pre-processing and augmentation.
  • Normalization and resizing for ResNet.

• Fast Deployment

  • Hosted on Hugging Face Spaces.
  • Runs efficiently in a lightweight environment.

---

5. Technology Stack

• Python 3.12+
• PyTorch & Torchvision
• Gradio for UI
• Hugging Face Hub for dataset & model storage

---

6. Project Structure

Smart_Waste_Classification_Project/
├── README.md
├── images/
│   ├── cardboard.jpg
│   ├── The_apps.png
│   ├── The_result.png
├── data/
│   ├── train/
│   ├── val/
├── model.pth
├── train.py
├── app.py
├── hf_upload_dataset_model.py
├── hf_deploy_space.py
├── requirements.txt
└── .env

---

7. Installing & Running

Step 1: Install Dependencies

pip install --upgrade huggingface_hub transformers requests tqdm torch torchvision gradio python-dotenv

Step 2: Set Up Hugging Face Token

Create a .env file and add:

HF_TOKEN=hf_XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Step 3: Upload Dataset & Model

python hf_upload_dataset_model.py

This uploads: ✅ Dataset → Smart-Waste-Dataset-Reduced
✅ Model → Smart-Waste-Classifier-ResNet18

---

8. Processing Pipeline

graph TD
    A[Image Input] --> B[Data Preprocessing]
    B --> C[Feature Extraction]
    C --> D[ResNet18 Model]
    D --> E[Classification Output]

Loading

1. Preprocess images (resize, normalize).
2. Extract features using ResNet18.
3. Classify into six categories.

---

9. Data Analysis

Sample Classification Result

Prediction: Cardboard (75%)
Top 3 Classes:
1. Cardboard - 75%
2. Paper - 21%
3. Trash - 2%

---

10. Deployment Steps

Deploy Gradio App on Hugging Face

python hf_deploy_space.py

This deploys to Hugging Face Spaces: ✅ Live App → Smart-Waste-Classifier-App

---

11. References

• Hugging Face Docs
• PyTorch Documentation
• Gradio Guide

---

12. License

Copyright (c) 2023 Fahmi Zainal

This project is licensed for personal and educational use only. Modification and redistribution require explicit permission from the author.

🚀 Enjoy using Smart Waste Classifier!