Melanoma Skin Cancer Detection Dataset Overview

Dataset Description

• Source: International Skin Imaging Collaboration (ISIC)
• Scale: 2,357 images of malignant and benign tumor lesions
• Categories: The dataset includes the following conditions:
  • Malignant melanoma
  • Benign melanoma
  • Other tumor diseases

Data Processing

• Class Balance: The Augmentor package was used for data augmentation to address class imbalance.

Data Examples

• Example Images: The dataset contains example images of skin lesions.

Model Architecture

• CNN Design:
  • Rescaling Layer: Scales input range from [0, 255] to [0, 1]
  • Convolutional Layer: Applies convolution to convert pixels to feature maps
  • Pooling Layer: Reduces feature map dimensions, decreasing parameters and computation
  • Dropout Layer: Randomly sets input units to zero to prevent overfitting
  • Flatten Layer: Flattens convolutional output to a 1‑D array
  • Dense Layer: Fully connected neural network layer
  • Activation Functions: ReLU and Softmax

Model Evaluation

• Evaluation Results: Shows the model's performance metrics and outcomes