Deploying a Pretrained Keras Model

• This code demonstrates how to deploy a Keras Model to Production.
• Its Subdivided into two main parts:

1. Backend - Traffic Signs Classifier as RESTful Service

   • Python Source Code in Root ./

2. Frontend - Consuming the RESTful service from a web app(react app).

   • React Source Code in Client Directory ./client

The trained model is available as a .h5 file name my_model.h5

The Source Code used to train the Model is available here: https://github.com/ItsCosmas/Traffic-Sign-Classification

API ENDPOINTS

• Production URL - https://damp-lake-03654.herokuapp.com
• Development URL - http://0.0.0.0:5000

- /classifier/ - accepts file.

• Example

1. cURL

curl --location --request POST 'http://0.0.0.0:5000/classifier/' \
--form 'file=@/path_to_file/image_filename.extension'

2. JavaScript

var formdata = new FormData();
formdata.append("file", fileInput.files[0], "image_filename.extension");

var requestOptions = {
  method: 'POST',
  body: formdata,
  redirect: 'follow'
};

fetch("http://0.0.0.0:5000/classifier/", requestOptions)
  .then(response => response.text())
  .then(result => console.log(result))
  .catch(error => console.log('error', error));

- /networkimg/ - accepts a JSON object containing an image url.

• Example Request

1. cURL

curl --location --request POST 'http://0.0.0.0:5000/networkimg/' \
--data-raw '{
	"image_url": "https://miro.medium.com/max/400/1*nhvFD7uT718W59UlRYaIWQ.jpeg"
}'

2. JavaScript

var raw = "{\n	\"image_url\": \"https://miro.medium.com/max/400/1*nhvFD7uT718W59UlRYaIWQ.jpeg\"\n}";

var requestOptions = {
  method: 'POST',
  body: raw,
  redirect: 'follow'
};

fetch("http://0.0.0.0:5000/networkimg/", requestOptions)
  .then(response => response.text())
  .then(result => console.log(result))
  .catch(error => console.log('error', error));