🍊 Citrus Leaf Disease Classification & Detection System

A high-performance Deep Learning application designed to classify and detect diseases in citrus leaves. This project utilizes the EfficientNetV2-S architecture with Transfer Learning techniques, achieving an impressive accuracy. It also features a user-friendly Desktop GUI for real-time inference on local images.

✨ Key Features

• State-of-the-Art Model: Built on EfficientNetV2-S architecture, offering an optimal balance between inference speed and classification accuracy.
• Two-Stage Training: Implements a robust training pipeline starting with Feature Extraction followed by Fine-Tuning strategies to maximize model performance.
• Hardware Optimized: Leverages Mixed Precision (BFloat16/Float16), Memory Format optimization (Channels Last), and RAM Caching to ensure fast and efficient training on NVIDIA GPUs.
• Smart Preprocessing: Includes HSV-based Leaf Segmentation algorithms to filter out background noise before prediction, improving real-world reliability.
• User-Friendly GUI: A clean and intuitive desktop interface built with PyQt6, allowing easy image loading and instant disease analysis.

📂 Project Structure

├── 📂 Dataset/                # Root directory for the image dataset
│   ├── Citrus_Canker/
│   ├── Black_Spot/
│   └── ...
├── 📂 Model/                  # Directory to store trained model weights
│   ├── checkpoint_fe.pth
│   └── model_final.pth
├── CitrusDetectorApp.py       # Main Python script for the Desktop GUI Application
├── train.ipynb                # Jupyter Notebook containing the full Training Pipeline
├── requirements.txt
└── README.md

🧠 Model & Classes

The project employs the EfficientNetV2-S (Small) architecture, trained using Transfer Learning. This specific model variant was selected for its superior trade-off between computational efficiency and accuracy, making it suitable for desktop deployment even on limited hardware.

The model is trained to classify citrus leaves into 5 specific categories:

1. Citrus Canker
2. Nutrient Deficiency
3. Multiple Diseases
4. Healthy Leaf
5. Young Healthy Leaf

Model Performance: The model achieved an accuracy of ~98.70% on the validation dataset after the Fine-Tuning stage.

---

🖥️ Development Environment

The model was trained and tested on the following system specifications:

⚙️ Hardware

Component	Specification
CPU	AMD Ryzen 5 7500F
GPU	NVIDIA GeForce RTX 5060
RAM	2 x 16GB 6000MHz CL30

💻 Software & Environment

Component	Version / Detail
OS	Windows 11 Pro 25H2 (26200.7462)
AMD	7.11.26.2142 (Chipset Driver)
NVIDIA	591.44 (Studio Driver)
CUDA	13.1 (NVCC), 9.17 (CUDNN)
Python	3.13.11 (pip 25.3)
PyTorch	2.9.1 (Stable CUDA 13.0)