🗑️ Waste Classification Using Convolutional Neural Network (CNN)

This project implements an image classification system to automatically classify waste into six categories using deep learning (CNN) with TensorFlow and Keras. The dataset used is TrashNet, which contains images of various types of waste.

📁 Project Structure

ML - TUBES/
├── Trashnet/                        # Original dataset (belum dipisah)
│   ├── cardboard/
│   ├── glass/
│   ├── metal/
│   ├── paper/
│   ├── plastic/
│   └── trash/
│
├── Trashnet_split/                 # Dataset setelah split: train/val/test
│   ├── train/
│   ├── val/
│   └── test/
│
├── ArielYosuaHasibuan_FP_ML2425_manual_book.pdf     # Manual book (penjelasan kode)
├── ArielYosuaHasibuan_FP_ML2425_paper.pdf           # Laporan akhir
├── ArielYosuaHasibuan_FP_ML2425_source_code.ipynb   # Notebook proyek
└── README.md                                        # Dokumentasi proyek (file ini)

🧾 Waste Categories

• cardboard
• glass
• metal
• paper
• plastic
• trash (minority class)

⚙️ Techniques Implemented

• ✅ Data Splitting: 70% train, 17% val, 13% test
• ✅ Image Preprocessing with ImageDataGenerator
• ✅ Data Augmentation focused on minority class trash
• ✅ Class Weighting to handle class imbalance
• ✅ Transfer Learning using pretrained MobileNetV2 and DenseNet121
• ✅ Fine-Tuning: Partial unfreezing of layers for performance improvement
• ✅ Callbacks:
  • ModelCheckpoint (save best model)
  • EarlyStopping (prevent overfitting)
  • ReduceLROnPlateau (adjust learning rate)

🧠 Model Architectures

Model	Notes
MobileNetV2	Lightweight, fast convergence
DenseNet121	Deeper model, higher final accuracy

📊 Performance Summary

Model	Accuracy	F1-Score	Recall (Trash)
MobileNetV2	~79%	0.77	0.67
DenseNet121	86%	0.86	0.83

📈 Evaluation

• Confusion matrix for each model
• Classification report: precision, recall, F1-score per class
• Accuracy and loss graphs (training vs validation)

🛠 How to Run

1. Clone this repository
   git clone https://github.com/arielyosua/Waste-Classification-Using-Convolutional-Neural.git cd Waste-Classification-Using-Convolutional-Neural

2. Install dependencies
   pip install -r requirements.txt

3. Make sure folder Trashnet is populated with original dataset images
   or download from TrashNet GitHub

4. Run source code in Jupyter Notebook or Python script
   jupyter notebook ArielYosuaHasibuan_FP_ML2425_source_code.ipynb

5. Output models will be saved as .h5 (MobileNetV2 and DenseNet121)

📄 Files Included

• *_manual_book.pdf → Detailed code explanation
• *_source_code.ipynb → Main training workflow
• *_paper.pdf → Formal report of the final project

👨‍💻 Author

This project was created by Ariel Yosua Hasibuan for the ML2425 - Final Project
Program Studi Ilmu Komputer, Universitas Pertamina
If you find this useful or want to contribute, feel free to fork or raise an issue.

---

⭐️ Don't forget to star the repository if you like it!