Waste Classification | Ojonugwa Wada

🧠 Overview

This project uses the EfficientNet-B0 model to classify waste into recyclable, organic, and other categories. It aims to assist automated waste separation systems to reduce landfill waste and improve environmental sustainability.

📂 Project Report & Notebook

🔗 View Full GitHub Repository
Includes complete Jupyter Notebook, training logs, and high-resolution visualizations

⚙️ Tools & Technologies

Python, TensorFlow, OpenCV, NumPy, Pandas
EfficientNet-B0 architecture
Data preprocessing, augmentation, transfer learning

📊 Model Performance

Test Accuracy: 94%

Evaluation on unseen waste images demonstrated the model's strong generalization capability with precision, recall, and F1-scores above 90% in each category.

💻 Sample Code Snippet

# Load EfficientNetB0 model
base_model = EfficientNetB0(include_top=False, input_shape=(224, 224, 3), weights='imagenet')

# Freeze the base
for layer in base_model.layers:
    layer.trainable = False

# Add custom classification head
model = tf.keras.Sequential([
    base_model,
    tf.keras.layers.GlobalAveragePooling2D(),
    tf.keras.layers.Dense(128, activation='relu'),
    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Dense(3, activation='softmax')
])

# Compile and train
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
model.fit(train_data, epochs=10, validation_data=val_data)

📈 Key Visualizations (Samples)

ℹ️ These visualizations show selected excerpts from the full analysis. For complete implementation details and interactive components:

Explore the Full Project Repository
Review the Jupyter Notebook with executable code
Access high-resolution charts and model architecture diagrams

1. Class Distribution

Balanced dataset across 3 waste categories (Recyclable, Organic, Other)

2. Image Dimensions

Original image size distribution before resizing to 224x224px

3. Dataset Split

70-15-15 split for training, validation, and testing sets

4. Training Progress

Convergence of training and validation metrics over epochs

5. Classification Report

Precision, recall, and F1-scores for each waste category

6. Confusion Matrix

Model prediction accuracy across different classes

7. Sample Predictions

Real-world classification examples with confidence scores