🧬 Melanoma Classification using CNN

This repository contains a Convolutional Neural Network (CNN) model for classifying melanoma from benign moles using dermoscopic images. The model is trained and evaluated on a dataset of skin lesion images.

🏗️ Model Architecture

The CNN architecture consists of the following layers:

Layer (type)	Output Shape	Param #
input_layer (InputLayer)	(None, 128, 128, 3)	0
conv2d (Conv2D)	(None, 128, 128, 32)	896
max_pooling2d (MaxPooling2D)	(None, 64, 64, 32)	0
batch_normalization (BatchNormalization)	(None, 64, 64, 32)	128
dropout (Dropout)	(None, 64, 64, 32)	0
conv2d_1 (Conv2D)	(None, 64, 64, 32)	9,248
max_pooling2d_1 (MaxPooling2D)	(None, 32, 32, 32)	0
batch_normalization_1 (BatchNormalization)	(None, 32, 32, 32)	128
dropout_1 (Dropout)	(None, 32, 32, 32)	0
flatten (Flatten)	(None, 32768)	0
dense (Dense)	(None, 512)	16,777,728
batch_normalization_2 (BatchNormalization)	(None, 512)	2,048
dropout_2 (Dropout)	(None, 512)	0
dense_1 (Dense)	(None, 512)	262,656
batch_normalization_3 (BatchNormalization)	(None, 512)	2,048
dropout_3 (Dropout)	(None, 512)	0
dense_2 (Dense)	(None, 2)	1,026

📁 Dataset

The dataset consists of dermoscopic images labeled as either "Benign" or "Malignant." Below is a sample visualization of the dataset:

📈 Training and Validation

The model was trained for 100 epochs. The performance metrics are shown in the plots below:

• Left Plot: Accuracy over epochs for training and validation sets.
• Right Plot: Loss over epochs for training and validation sets.

🏆 Results

The model achieved the following results:

• Training Accuracy: High accuracy with low loss.
• Validation Accuracy: Fluctuations indicate potential overfitting, which can be addressed by further tuning hyperparameters or augmenting the dataset.
• Testing Accuracy: Obtained 90.43% accuracy.

⚙️ Usage

Prerequisites

Ensure you have the following libraries installed:

pip install numpy opencv-python tensorflow matplotlib pillow scikit-learn

Running the Model

1. Clone the repository:

git clone <repository-url>
cd <repository-folder>

2. Run the model.ipynb code. It will save the trained model as an .h5 file named melanoma_cnn.h5 in your local Desktop folder.

3. Once the .h5 file is generated, upload your own image to the test_data folder. Then, edit cell 3 of model_test.ipynb to match the name of your file and run all the cells.

image = cv2.imread('test_data/name_of_your_image.jpg')

🌟 Future Work

• Implement data augmentation to improve generalization.
• Experiment with different architectures and hyperparameter tuning.
• Incorporate transfer learning for improved performance.