Dermdetect is a project to identify skin cancers from images. Using the Inception_v3 model and developed with Streamlit, this application allows you to upload images of skin lesions and get fast and accurate predictions.
- MLFlow: Used for lifecycle management of machine learning models.
- Streamlit: Used to build the web UI.
- Tensorflow: Used to implement the inception_v3.
- pandas: For data manipulation.
- plotly: For creating interactive visualizations.
- S3: To save artifacts
git clone https://github.com/nprevost/dermdetect.git
cd dermdetectpython3.12 -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
Copy the .env.tmp file, rename to .env and fill in the info.
- First you need to create, put the zip and unzip it in an images folder of the project.
- Then create a csv with the name of the images and the target. Group all the images in the same folder and retrieve the information with the ISIC API. To do this, launch preprocessing.py and clean_metadata.py which is in the scripts folder.
You need to create a database with neon db
Also create an S3 bucket
- In the streamlit folder, create a model folder and add the model.keras file to it
- Add the merge_metadata.csv in the S3 and change the value of url_csv in the dataset.py file
- In the streamlit folder, run the following commands to launch the docker
docker build . -t dermdetect_streamlit
docker run -it -v "$(pwd):/home/app" -e PORT=80 -p 4000:80 dermdetect_streamlit
In the .env file, add the link for MLFlow (Example):
APP_URI_MLFLOW=https://nprevost-dermdetect-mlflow.hf.space
To retrieve them in the python script
load_dotenv(dotenv_path='.env')
csv_path = os.getenv('CSV_PATH')
dermdetect/
├── .venv/ # Environment python
├── benchmark_models/ # Model used to analyze performance
│ ├── artefacts/ # Models and matrix saved
│ ├── baseline_cnn.py # Model base CNN
│ ├── resnet50.py # Model resnet50
│ ├── vgg16.py # Model VGG16
│ └── inceptionv3.py # Model inception_v3
├── csv/ # Datasets
├── images/ # Images for training
├── mlflow/ # Files for mlflow
│ ├── Dockerfile # Docker configuration file
│ ├── README.md # Config MLFlow
│ └── requirements.txt # Dependencies for MLFlow
├── model/ # Best model
├── scripts/ # Scripts python
│ ├── clean_metadata.py # Scripts to create clean metadata
│ ├── preprocessing.py # Scripts to create dataset and move images
│ └── script_test_mlflow.py # Small script to test the connection to mlflow
├── streamlit/ # Files for streamlit
│ ├── .streamlit/
│ │ └── config.toml # Config streamlit
│ ├── model/ # Best model to predict in streamlit
│ ├── pages/ # Different pages of the streamlit
│ | ├── dataset.py # Page to analyze the dataset
│ │ ├── intro.py # Streamlit Home Page
│ | └── model.py # Page to make the prediction
│ ├── Dockerfile # Docker configuration file
| |── requirements.txt # Dependencies for streamlit
│ └── streamlit_app.py # Menu and page management
├── .env # Environment variables local
├── .env.tmp # Environment variables Example
├── .gitignore # Git ignore file
├── README.md # Project documentation
└── requirements.txt # Dependencies for the project