🌿 LeafLens — AI Plant Disease Classifier

Upload a leaf image. Get an instant disease diagnosis powered by deep learning + expert LLM advice for farmers.

---

📌 What It Does

LeafLens is an end-to-end plant disease detection web app that:

• Classifies leaf images into 38 disease/healthy categories across 14 crop types
• Shows top-3 predictions with confidence scores
• Provides AI-generated expert advice via LLM (Groq + LLaMA 3.1) — symptoms, prevention, and treatment in simple farmer-friendly language

---

🧠 Model Comparison

Three architectures were benchmarked on the PlantVillage dataset before selecting the final model:

Model	Val Accuracy	Params	Training Speed
ResNet18	~93.2%	11.7M	Fast
DenseNet121	~95.8%	8.0M	Moderate
EfficientNet-B0 ✅	98.65%	5.3M	Fast

EfficientNet-B0 was chosen — highest accuracy with the fewest parameters, making it efficient for deployment without sacrificing performance.

Accuracy & Loss Curves

Confusion Matrix

---

🤖 LLM Integration

After classification, the top-3 predictions are passed to LLaMA 3.1-8B via Groq API which generates:

1. Disease explanation
2. Symptoms to look for
3. Prevention methods
4. Treatment options

All explained in simple language suitable for farmers — not just a label, but actionable advice.

---

🌱 Supported Crops & Diseases (38 Classes)

Crop	Conditions Covered
Tomato	Late blight, Early blight, Bacterial spot, Leaf mold, Mosaic virus, Yellow Leaf Curl, Septoria leaf spot, Spider mites, Target spot, Healthy
Apple	Apple scab, Black rot, Cedar apple rust, Healthy
Grape	Black rot, Esca (Black Measles), Leaf blight, Healthy
Corn	Northern Leaf Blight, Cercospora leaf spot, Common rust, Healthy
Potato	Late blight, Early blight, Healthy
Peach	Bacterial spot, Healthy
Cherry	Powdery mildew, Healthy
Strawberry	Leaf scorch, Healthy
Pepper	Bacterial spot, Healthy
Orange	Haunglongbing (Citrus greening)
Squash	Powdery mildew
Soybean	Healthy
Blueberry	Healthy
Raspberry	Healthy

---

🛠️ Tech Stack

Component	Technology
Model	EfficientNet-B0 (PyTorch, Transfer Learning)
Dataset	PlantVillage (Kaggle) — 87,000+ images
Frontend	Streamlit
LLM	LLaMA 3.1-8B via Groq API
LLM Framework	LangChain OpenAI

---

🚀 Run Locally

1. Clone the repo

git clone https://github.com/TheGhostLoop/leaflens.git
cd leaflens

2. Install dependencies

pip install -r requirements.txt

3. Add your Groq API key

Create .streamlit/secrets.toml:

GROQ_KEY = "your_groq_api_key_here"

Get a free key at console.groq.com

4. Run the app

streamlit run app.py

---

📁 Project Structure

leaflens/
├── app.py                  # Main Streamlit app
├── model.pth               # Trained EfficientNet-B0 weights
├── requirements.txt
├── accuracy_graph.png      # Training vs validation accuracy
├── confusion_matrix.png    # Model evaluation
├── .gitignore
└── .streamlit/
    └── secrets.toml        # Local only — never pushed

---

📦 Requirements

streamlit
torch
torchvision
Pillow
langchain-openai
pandas

---

📊 Dataset

PlantVillage — publicly available on Kaggle

• 87,000+ leaf images
• 38 classes (healthy + diseased)
• 14 crop types
• RGB images, varied lighting and backgrounds

---

👤 Author

Prince Verma

• GitHub: @TheGhostLoop
• Email: princeverma2005@yahoo.com

---

📄 License

This project is open source and available under the MIT License.