Brain Tumor Classification Using MobileNet and TensorFlow

Overview

This project implements a deep learning pipeline for classifying brain MRI images into four categories:

• Glioma
• Meningioma
• Pituitary tumor
• No tumor

The model leverages MobileNet as a pre-trained feature extractor with a custom classifier on top.
It uses TensorFlow 2.10.1, runs on GPU if available, and includes data augmentation, early stopping, and model checkpointing.

---

Features

• Preprocessing and loading images from directory using tf.keras.utils.image_dataset_from_directory
• Data augmentation: Random brightness and contrast adjustments
• Transfer learning with MobileNet (pre-trained on ImageNet)
• Custom fully connected classifier layers with BatchNormalization and Dropout
• Early stopping and best model checkpoint saving
• GPU acceleration support via CUDA/cuDNN (if available)
• Evaluation with accuracy, confusion matrix, and classification report
• Visualization of training/validation accuracy and loss
• Visualization of predictions on test images

---

Requirements

• Python 3.10
• TensorFlow 2.10.1
• NumPy 1.26.4
• Pandas
• Matplotlib
• Seaborn
• Scikit-learn
• CUDA 11.2 and cuDNN 8.1 (for GPU support)
• Conda (recommended for environment management)

---

Directory Structure

project_root/
│
├─ data/
│ ├─ train/
│ │ ├─ glioma/
│ │ ├─ meningioma/
│ │ ├─ notumor/ │ │ └─ pituitary/
│ └─ test/
│ ├─ glioma/
│ ├─ meningioma/
│ ├─ notumor/
│ └─ pituitary/ │
├─ BrainTumourDetection.ipynb # Jupyter notebook with model training & evaluation
├─ best_model_mobileNet.keras # Saved best model weights
└─ README.txt