Week 2 – Smart Waste Classification -- CNN Model Building & Training
♻️ Smart Waste Classification – Week 2
🧠 AICTE × Shell Edunet Foundation – Sustainability Internship
🎯 Objective
The goal of Week 2 is to design, train, and evaluate a Convolutional Neural Network (CNN) model for classifying waste images into recyclable and non-recyclable categories.
This project supports sustainable waste management by automating the segregation process using deep learning.
📦 Dataset
Source: Waste Classification Data (Kaggle)
Classes:
- 🧃 Cardboard
- 🍾 Glass
- 🧲 Metal
- 📄 Paper
- 🛍️ Plastic
- 🗑️ Trash
Structure: Waste Classification Data/ ├── TRAIN/ │ ├── cardboard/ │ ├── glass/ │ ├── metal/ │ ├── paper/ │ ├── plastic/ │ └── trash/ └── TEST/
⚙️ Steps Performed
1️⃣ Data Loading
- Loaded dataset using
ImageDataGenerator - Split data: 80 % training / 20 % validation
2️⃣ Model Building
A Sequential CNN with:
- Conv2D → ReLU → MaxPooling (×3 blocks)
- Flatten → Dense(128, ReLU) → Dropout(0.5)
- Output Dense(6, Softmax)
3️⃣ Model Compilation
- Optimizer: Adam
- Loss: Categorical Crossentropy
- Metrics: Accuracy
4️⃣ Training & Evaluation
- Epochs: 10
- Batch Size: 32
- Validation Split: 0.2
- Trained on GPU (Google Colab)
5️⃣ Model Saving & Visualization
- Saved model →
waste_cnn_model.h5 - Plotted accuracy curve →
accuracy_plot.png
📊 Results
| Metric | Value |
|---|---|
| Training Accuracy | ~87 % |
| Validation Accuracy | ~83 % |
| Loss Function | Categorical Crossentropy |
| Optimizer | Adam |
| Epochs | 10 |
📈 Accuracy Curve
🧩 Model Summary : Total Parameters: 1,034,182 Trainable Parameters: 1,034,182 Input Shape: (128, 128, 3) Output Classes: 6
🧠 Tools & Libraries
- TensorFlow / Keras
- NumPy / Matplotlib
- Google Colab
- Kaggle Dataset
🌍 Sustainability Impact
This model contributes to the UN Sustainable Development Goals:
- SDG 11 – Sustainable Cities & Communities
- SDG 12 – Responsible Consumption & Production
- SDG 13 – Climate Action
By automating waste classification:
- Reduces manual sorting errors
- Improves recycling efficiency
- Encourages eco-friendly disposal practices
📁 Repository Structure
Week 2/
├── week2_model_training.ipynb # CNN model training notebook
├── waste_cnn_model.h5 # Saved model file (https://drive.google.com/file/d/12niVb4fBtlwc7gEuJ5XPXTCHl35DPHd-/view?usp=drive_link)
├── accuracy_plot.png # Accuracy graph
├── dataset_link.txt # Kaggle dataset reference
└── README.md
Author
Salla Prathik Reddy
AICTE × Shell Edunet Foundation – Sustainability Internship
Intern ID: INTERNSHIP_175683301568b724f7b9fba
Week 2 Milestone – Model Building & Training Completed