This project provides an end-to-end solution for kidney disease classification using machine learning. It leverages MLflow for experiment tracking and model management, as well as DVC for data versioning and pipeline management.
In this project, we have utilized the VGG16 convolutional neural network architecture pretrained on the ImageNet dataset. We fine-tuned this model on our kidney disease dataset to create a custom classifier.
To install the necessary dependencies, run the following command:
pip install -r requirements.txtMake sure you have Python 3.6 or later installed.
- First, clone the repository:
git clone https://github.com/Aftabmallick/End-to-End-Kidney-Disease-Classification-With-MLFLOW-and-DVC.git- Navigate to the project directory:
cd End-to-End-Kidney-Disease-Classification-With-MLFLOW-and-DVC- Run the Jupyter notebook to train the model and evaluate results:
jupyter notebook- Alternatively, you can run the Flask web application for inference:
python app.py- To retrain the model using DVC, use the dvc repro command:
dvc reproThis command will reproduce the pipeline defined in the dvc.yaml file, which includes data preprocessing, model training, and evaluation. Any changes to the data or code will trigger DVC to rerun the pipeline, ensuring reproducibility.
We have retrained the VGG16 model using our kidney disease dataset. The pretrained weights from ImageNet were fine-tuned during this process. The retrained model achieves state-of-the-art performance on our task.
- tensorflow==2.12.0
- pandas
- dvc
- mlflow==2.2.2
- notebook
- numpy
- matplotlib
- seaborn
- python-box==6.0.2
- pyYAML
- tqdm
- ensure==1.0.2
- joblib
- types-PyYAML
- scipy
- Flask
- Flask-Cors
- gdown
You can install these dependencies using the provided requirements.txt file.
- This project was inspired by the need for a comprehensive solution for kidney disease classification.
- Special thanks to the contributors of the dependencies used in this project.