Welcome to the MoleSafeScan App repository! This app is a tool for classifying skin lesions as benign or malignant using a convolutional neural network (CNN) based on the MobileNetV2 architecture. It is built with Streamlit for easy web deployment and TensorFlow for constructing and training the model.
This app is intended for educational and research purposes only and is not meant to be a substitute for professional medical advice or diagnosis.
- Upload multiple images of skin lesions for classification.
- Classify lesions with a pre-trained MobileNetV2 model fine-tuned on a proprietary skin lesion dataset.
- View classification results directly in the app with confidence scores.
- Architecture: MobileNetV2 pre-trained on ImageNet.
- Fine-tuning: Started from layer 154 of MobileNetV2.
- Test Accuracy: 91.04%.
- Dataset: Proprietary dataset containing images of benign and malignant skin lesions.
To run the app, you'll need:
- Python 3.7+
- Streamlit
- TensorFlow 2.15.0
- NumPy
- Pillow (PIL)
-
Clone the repository to your local machine.
-
Install the required Python packages:
pip install streamlit tensorflow==2.15.0 numpy pillow -
Run the app:
streamlit run app.py
The model was trained in a Jupyter notebook (Skin_Cancer.ipynb) with the following key steps:
- Data preprocessing and augmentation using
ImageDataGenerator. - Model definition with TensorFlow's Keras API.
- Fine-tuning on a skin lesion dataset from layer 154.
For detailed training procedures, hyperparameter tuning, and learning rate analysis, refer to the Jupyter notebook.
After training, the model is saved and can be converted to TensorFlow Lite format for deployment in mobile applications. You can test the model here : https://psilly-billy-molesafescan.streamlit.app
Contributions to the app or model are welcome. Please submit a pull request or open an issue if you have suggestions or find a bug.
Distributed under the MIT License. See LICENSE for more information.
"""