aziz00legend/AI-Degla

A deep learning project focused on classifying 11 Tunisian date varieties using Convolutional Neural Networks (CNNs), combining advanced image enhancement, data augmentation, and explainable AI techniques for accurate and interpretable results. # 🌴 Tunisian Date Classification AI Project ## Project Overview This project develops an advanced image classification system to identify and classify 11 different varieties of Tunisian dates using deep learning techniques. The system leverages Convolutional Neural Networks (CNNs) to distinguish between date varieties based on their unique visual characteristics. ## 🎯 Objectives - Develop a high-performing CNN model for Tunisian date variety classification - Implement advanced image processing and enhancement techniques - Achieve robust model performance through cross-validation and data augmentation - Provide explainable AI insights using LIME for model interpretability - Support agricultural stakeholders with accurate date variety identification ## 📊 Dataset **Source:** Deglet Nour Date Fruit Dataset **Classes:** 11 different Tunisian date varieties **Size:** 300-473 images per class ### Date Varieties Included: - Bessra - Deglet Nour Dryer - Deglet Nour Oily - Deglet Nour Semi-Oily Treated - Deglet Nour Semi-Dryer - Alig - Kenta - Deglet Nour Oily Treated - Deglet Nour Semi-Dryer Treated - Deglet Nour Semi-Oily - Deglet Nour ## 🔧 Technical Architecture ### Model Architecture - **Input Layer:** 224x224x3 RGB images - **Convolutional Layers:** Multiple Conv2D layers with ReLU activation - **Pooling Layers:** Combination of AveragePooling2D and MaxPooling2D - **Dense Layers:** Fully connected layers with dropout regularization - **Output Layer:** 11-class softmax classification ### Image Enhancement Pipeline 1. **Resizing:** Standard 224x224 pixel dimensions 2. **Normalization:** Pixel value normalization for improved contrast 3. **Sharpening:** Convolutional filter for edge enhancement 4. **CLAHE:** Contrast Limited Adaptive Histogram Equalization for local contrast enhancement ## 🚀 Methodology Following **CRISP-DM** methodology: ### 1. Business Understanding - Define classification objectives and success criteria - Identify stakeholder requirements (f …