Blur or Bokeh - Company Capstone Bangkit

Authored by Arkan Alexei Andrei and Maria Aurellia

Setup

pip install -r requirements.txt

Then run the app using python app.py

Feel free to use the S or L version of the model. Loading the L model might take a while.

Sample Usage

Image File

Endpoint: /predict

Sample Request

curl -X POST -F "files=@pictures/normal.jpg" -F "files=@pictures/bokeh.jpg" -F "files=@pictures/blur.png" http://localhost:8080/predict

Sample Response

[
  {
    "filename": "normal.jpg",
    "prediction": "Normal"
  },
  {
    "filename": "bokeh.jpg",
    "prediction": "Bokeh"
  },
  {
    "filename": "blur.png",
    "prediction": "Blur"
  }
]

Image URL

Endpoint: /predict-url

Sample Request

{
    "urls": [
        "https://st5.depositphotos.com/55662820/62581/i/450/depositphotos_625819662-stock-photo-blurred-purpose-urban-traffic-lights.jpg",
        "https://photographylife.com/wp-content/uploads/2009/12/Creamy-Bokeh.jpg",
        "https://sussexbylines.co.uk/wp-content/uploads/2024/03/cat-out-hunting.jpg"
    ]
}

Sample Response

[
    {
        "prediction": "Blur",
        "url": "https://st5.depositphotos.com/55662820/62581/i/450/depositphotos_625819662-stock-photo-blurred-purpose-urban-traffic-lights.jpg"
    },
    {
        "prediction": "Bokeh",
        "url": "https://photographylife.com/wp-content/uploads/2009/12/Creamy-Bokeh.jpg"
    },
    {
        "prediction": "Normal",
        "url": "https://sussexbylines.co.uk/wp-content/uploads/2024/03/cat-out-hunting.jpg"
    }
]

Deployment

https://capstone-ml-app-mo5jvyk6cq-as.a.run.app/

Rebuilding

docker build -t gcr.io/bangkit-capstone-ml/capstone-ml-app .

docker push gcr.io/bangkit-capstone-ml/capstone-ml-app

gcloud run deploy capstone-ml-app --image gcr.io/bangkit-capstone-ml/capstone-ml-app --platform managed --region asia-southeast1 --allow-unauthenticated --memory=2Gi --cpu=2

Note: deploying with 2GiB memory and 2 CPU seems to work well. Didn't work with 1 GiB memory and less.