Welcome to the Brain Tumor Classification project, where we have harnessed the power of deep learning to classify brain MRI images into three categories: meningioma, glioma, and pituitary tumors. Our project is also available on Kaggle at this link.
- Deep Learning Model: We've developed a deep learning model using PyTorch and the ResNet50 architecture to accurately classify brain MRI images.
- Data Preprocessing: Prior to training our model, we preprocessed the MRI images using Python, NumPy, and OpenCV. This preprocessing involved noise removal and normalization of pixel values, ensuring the model receives clean and consistent input data.
- Impressive Accuracy: Our model achieved an accuracy of 95% in classifying brain tumors, providing reliable results for medical professionals and researchers.
- K-Fold Cross-Validation: To ensure the robustness of our model, we implemented k-fold cross-validation, which helps mitigate overfitting and ensures generalizability.
- Mean Test Accuracy: The mean test accuracy of our model is 0.9434, with a low variance of +/- 0.0111, indicating the consistent and reliable performance of our model.
You can also find our project on Kaggle, where we provide code and additional resources related to brain tumor classification. Visit our Kaggle repository here for further details.
In this GitHub repository, you'll find the following:
- The deep learning model's code and architecture.
- Data preprocessing scripts.
- Training and evaluation scripts.
- Model weights and checkpoints.
- Documentation and instructions on how to use and evaluate the model.
- Sample brain MRI images for testing.
To utilize our model for brain tumor classification, follow the instructions in our documentation. We've provided clear steps to help you get started with classifying brain MRI images effectively.
Contributions to this project are welcome on both Kaggle and GitHub! Whether you're interested in improving the model's performance, adding new features, or enhancing the documentation, we encourage your participation. Please feel free to submit pull requests and report any issues you encounter.