🧬 Blood Cell Cancer Detection using CNN & EfficientNetB3

📌 Overview

This project focuses on detecting blood cell cancer (leukemia) using deep learning models. It utilizes a custom Convolutional Neural Network (CNN) and EfficientNetB3 (transfer learning) to classify microscopic blood cell images as cancerous or healthy.

The goal is to assist medical diagnosis by automating image-based detection, improving both speed and accuracy.

---

🎯 Objectives

• Build a deep learning model for blood cell classification
• Compare performance between CNN and EfficientNetB3
• Apply transfer learning for improved accuracy
• Create an end-to-end image classification pipeline

---

📂 Dataset

• Microscopic blood cell images
• Binary classification: Cancerous / Non-cancerous

Preprocessing Steps:

• Image resizing
• Normalization
• Data augmentation (rotation, flipping, zoom)

---

⚙️ Tech Stack

• Language: Python
• Libraries: TensorFlow, Keras, NumPy, Matplotlib, OpenCV
• Models: CNN, EfficientNetB3

---

🧠 Model Architecture

🔹 CNN Model

• Convolutional + MaxPooling layers
• Fully connected Dense layers
• Dropout for regularization

🔹 EfficientNetB3

• Pretrained on ImageNet
• Fine-tuned for classification
• Better feature extraction and generalization

---

🚀 Workflow

1. Data Loading
2. Data Preprocessing & Augmentation
3. Model Building (CNN & EfficientNetB3)
4. Model Training
5. Evaluation
6. Prediction

---

📊 Results

• EfficientNetB3 achieved higher accuracy than CNN
• Transfer learning significantly improved performance
• Reduced overfitting using augmentation and dropout