DermatoVisionAPI

DermatoVisionAPI is a RESTful API for dermatological image analysis. It focuses on predicting whether a skin condition is cancerous or not based on input images.

Features

• Cancer Detection: Given an input image, the API predicts whether the skin condition is cancerous or not.
• Model Storage: The API provides functionality to load and manage the machine learning model used for prediction.
• Input Validation: Input images are validated to ensure they meet the necessary criteria before prediction.

Prerequisites

• Node.js (v20.12.2)
• Google Cloud SDK
• Docker
• A Google Cloud Project

Local Development

1. Clone the repository:

   git clone https://github.com/chrystalio/DermatoVisionAPI.git

2. Change the directory:

   cd DermatoVisionAPI

3. Install dependencies:

   npm install

4. [OPTIONAL] Upload your model to a Google Cloud Storage bucket. The model files are inside the model directory.

5. If you prefer to load the model data from the local directory, you can refer to it directly from the model directory.

6. Create a new service account in Google Cloud and assign the Cloud Datastore Owner role.

7. Download the JSON key for the service account, store it inside the config folder, and rename it to service.json.

8. Ensure your config/service.json contains your Google Cloud credentials.

9. Enable Firestore API.

10. Open Google Cloud Console → Firestore, then click Create Database.

11. Choose Native Mode, select your region, and create the database.

12. Set up your environment variables. Create a .env file in the root directory and add the following:

    APP_ENV=local
    MODEL_URL=https://storage.googleapis.com/your-bucket/model.json
    GCLOUD_PROJECT_ID={projectID}
    GCLOUD_KEY_FILE=config/service.json
    

    Or if you are not uploading the model to a cloud storage bucket, set it like this:

    APP_ENV=local
    MODEL_URL=./model/model.json
    GCLOUD_PROJECT_ID={projectID}
    GCLOUD_KEY_FILE=config/service.json
    

13. Start the server:

    npm run start:dev

Deploying on Google Cloud Platform

1. Build the Docker image:

   docker build -t {region}-docker.pkg.dev/{projectID}/dermatovision-image/dermatovision-api:v1 .

2. Create Docker Repository inside Artifact Registry

   gcloud artifacts repositories create dermatovision-image --repository-format=docker --location={region} --description="DermatovisionAPI Docker Repository" --project={projectID}

3. Authenticate Docker to use the Artifact Registry repository

   gcloud auth configure-docker {region}-docker.pkg.dev

4. Push the Docker image to the Artifact Registry repository

    docker push {region}-docker.pkg.dev/{projectID}/dermatovision-image/dermatovision-api:v1

5. Deploy it on Cloud Run

   gcloud run deploy DermatoVisionAPI --image={region}-docker.pkg.dev/{projectID}/dermatovision-image/dermatovision-api:v1 --region={region} --allow-unauthenticated

Troubleshooting

If you encounter issues with permissions or environment variables, ensure that:

Your Google Cloud credentials are correctly set up in the config/credentials.json file. The environment variables in the .env file are correctly configured. The IAM roles for your service account have the necessary permissions.

Acknowledgements

• Thanks to the Dicoding team for providing the machine learning model.
• Special thanks to the Google Cloud Platform for providing the infrastructure.

Make sure to replace placeholder values (like `{projectID}`, `your-bucket`) with the actual values for your project. Additionally, include any other specific setup instructions or dependencies unique to your project.