This project aims to develop a deep learning model for the detection of skin cancer from dermoscopic images. The model utilizes convolutional neural networks (CNNs), specifically the ResNet50 architecture, to classify images into two classes: benign and malignant.
The dataset used in this project is the Skin Cancer dataset, obtained from Kaggle. It consists of dermoscopic images of skin lesions, categorized into benign and malignant classes.
data/train/: Training images divided into benign and malignant classes.validation/: Validation images for model evaluation.test/: Test images for final model evaluation.
model_version_*_log.csv: CSV files containing training logs for each model version.model_version_*.h5: Saved weights of each trained model version.resNet50_V3.h5: Final trained model.
- Setup: Install the necessary libraries and download the dataset from Kaggle.
- Preprocessing: Balance the dataset and divide it into training, validation, and test sets.
- Modeling: Train the model using the ResNet50 architecture, repeating the process 6 times to account for variability.
- Evaluation: Evaluate the model on the test set and analyze performance metrics such as accuracy, precision, recall, specificity, and sensitivity.
- Testing: Test the final model on unseen data.
- Implement techniques such as SMOTE and data augmentation to improve model performance.
- Experiment with more complex architectures and ensemble methods.
- Continuously monitor and update the model to improve its accuracy and generalization.
- Alphabetically:
This project is licensed under the License.