🌿 Plant Disease Classification using ResNet50

This project uses a deep learning model (ResNet50) to classify 38 different plant diseases using leaf images from the PlantVillage dataset.

---

📁 Dataset

• Source: PlantVillage (Augmented)
• Number of Classes: 38
• Image Size: 224x224
• Directory Structure:
  • train/
  • valid/

---

🧠 Model

• Base Model: ResNet50 (pretrained on ImageNet)
• Custom Layers:
  • GlobalAveragePooling
  • Dropout (0.3)
  • Dense (128, ReLU)
  • Dropout (0.3)
  • Dense (Softmax - 38 classes)
• Training Strategy:
  • Phase 1: Base model frozen (15 epochs)
  • Phase 2: Fine-tuning (20 epochs)
• Optimizer: Adam
• Loss Function: Categorical Crossentropy
• Final Validation Accuracy: 99.69%

---

📊 Results

• 🟦 Confusion Matrix: results/confusion_matrix.jpg
• 📈 Accuracy & Loss Curves: results/training_curves.jpg
• 📃 Full classification report: see notebooks/evaluation.ipynb

---

🚀 How to Use

1. Clone this repository:

   git clone https://github.com/YOUR_USERNAME/plant-disease-resnet50.git
   cd plant-disease-resnet50

2. Install the required libraries:

   pip install -r requirements.txt

3. Download the trained model (see below), and place it inside the model/ directory.

4. Run the Jupyter notebooks in order:

   • notebooks/data_preparation.ipynb
   • notebooks/train_resnet50.ipynb
   • notebooks/evaluation.ipynb

---

📦 Requirements

Install dependencies using the command:

pip install -r requirements.txt

---

🔗 Download Pretrained Model

The trained model exceeds GitHub’s file size limit (100MB).
You can download it directly from Google Drive:

📥 Download Plant_38_model.keras

After downloading, place the file here:

plant-disease-resnet50/model/Plant_38_model.keras

---

👨‍💻 Author

Hüseyin Oso
📧 Email: hosenoso25@gmail.com
🌐 GitHub: HossenOSO