🏆 CNN-Based Image Classification

🚀 Supervised Learning with Deep Neural Networks using Transfer Learning (InceptionV3)

📌 Project Overview

This project focuses on lung disease classification using Convolutional Neural Networks (CNNs) and Transfer Learning.
It applies InceptionV3 to classify lung X-ray images into three categories:
1️⃣ COVID-19 Affected Lungs
2️⃣ Pneumonia Affected Lungs
3️⃣ Normal Lungs

📂 Dataset

• Total Images:
  • Training Set: 10,606
  • Validation Set: 3,030
  • Test Set: 1,517
• Number of Classes: 3 (COVID, Pneumonia, Normal)
• Image Dimensions: Varying sizes (e.g., 300×225px)

🛠️ Methodology

✅ Preprocessing: Resizing, Normalisation, Data Augmentation
✅ Model Architecture: InceptionV3 (Pretrained on ImageNet)
✅ Fine-Tuning: Adjusting layers for optimal performance
✅ Optimisation: Adam Optimiser with Categorical Crossentropy Loss
✅ Evaluation Metrics: Accuracy, Confusion Matrix, Classification Report

📊 Results & Performance

• Best Accuracy Achieved: 94%
• Batch Size Tested: 32, 64, 128 (Best Performance)
• Dropout Rate Tested: 0.3 – 0.5 (Best Stability)

🚀 How to Run

1️⃣ Clone the repository

git clone https://github.com/sv3112/CNN-Image-Classification.git
cd CNN-Image-Classification

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Open the Jupyter Notebook

jupyter notebook Lung_Dataset_Image_Classification.ipynb


📌 Future Improvements

🏆 Try ResNet50 or EfficientNet for comparison
📊 Implement Grad-CAM for Explainability
☁️ Deploy the model using Flask/FastAPI on AWS/GCP