The Skin Cancer Detection Project aims to evaluate and compare the performance of four deep learning models—CNN, MobileNetV3 Large, VGG16, and EfficientNetB4—for the automated detection and classification of skin lesions. The goal is to determine the most effective model for early skin cancer diagnosis, improving healthcare.
- Assess the effectiveness of each deep learning algorithm in detecting skin lesions.
- Compare model accuracy, training efficiency, and classification performance.
- Implement a CNN alongside pre-trained models (MobileNetV3 Large, VGG16, EfficientNetB4).
- Utilize transfer learning to fine-tune pre-trained models for skin cancer classification.
- Train models using TensorFlow and Keras in Google Colab.
- Develop a website linked to the trained model.
- Assist doctors in making informed decisions by providing automated skin cancer detection.
- The Skin Cancer Detection Project utilizes the skin-cancer9-classesisic dataset, which comprises 10,015 dermatoscopic images categorized into nine classes of skin lesions.
- For more information and access to the dataset, you can visit dataset page on Kaggle.
- Dataset URL: Skin Cancer 9 Classes (ISIC)
- The dataset consists of labeled images of skin lesions categorized into nine different classes.
- It includes high-resolution dermatoscopic images collected from public medical image datasets.
- Images are preprocessed using normalization, data augmentation, and resizing for optimal model performance.
- The dataset is split into training, validation, and testing sets to ensure proper evaluation and generalization.
- Achieved the highest validation accuracy of 95.78%.
- Showed consistent improvement in accuracy over 25 epochs.
- Utilized early stopping to prevent overfitting.
- Achieved a final accuracy of 89.47%.
- Improved steadily in performance over training epochs.
- Demonstrated good generalization to unseen data.
- Achieved a validation accuracy of 86.39%.
- Stopped training early at epoch 21 to prevent overfitting.
- Demonstrated strong performance in classification tasks.
- Achieved a peak validation accuracy of 87.56%.
- Implemented early stopping to maintain optimal performance.
- Showed a clear learning curve across 25 epochs.
- Programming Language: Python
- Deep Learning Framework: TensorFlow, Keras
- Libraries: NumPy, Pandas, Matplotlib, Seaborn, Scikit-learn
- Development Environment: Google Colab
- Clone the repository:
- Navigate to the project directory:
cd Skin-Cancer-Detection - Install dependencies:
pip install -r requirements.txt
- Run the application:
python app.py
