How to get prediction from ml rest api if it is an image classification model by keras for real time endpoint?

Isaac Tong 0

Background: Model is deployed to an endpoint and it's working. I am trying to get a prediction from the model by submitting an image.

My understanding is Keras model takes images as an array. However, converting an image to array would return a large array and it needs to be passed in the below code as a json body.

headers = {'Content-Type':'application/json', 'Authorization':('Bearer '+ api_key), 'azureml-model-deployment': 'blue' }

url = 'https://xyz.xyz.inference.ml.azure.com/score'
req = urllib.request.Request(url, body, headers)

try:
    response = urllib.request.urlopen(req)

    result = response.read()
    print(result)
except urllib.error.HTTPError as error:
    print("The request failed with status code: " + str(error.code))

    # Print the headers - they include the requert ID and the timestamp, which are useful for debugging the failure
    print(error.info())
    print(error.read().decode("utf8", 'ignore'))

I have tried adding "GlobalParameters": 1.0 but it did not work at all.

Error message :
The request failed with status code: 424server: azureml-frontdoordate: Fri, 24 Mar 2023 05:59:11 GMTcontent-type: application/jsoncontent-length: 38x-ms-run-function-failed: Truex-ms-server-version: azmlinfsrv/0.8.0x-ms-request-id: 34201170-065b-43af-97ea-7b76fa7ff2ffx-request-id: 34201170-065b-43af-97ea-7b76fa7ff2ffms-azureml-model-error-reason: model_errorms-azureml-model-error-statuscode: 500azureml-model-deployment: blueconnection: close

Ramr-msft 17,611 Reputation points

2023-03-24T12:41:07.09+00:00

@Isaac Tong . It would be helpful to check the logs for more detailed information about the error and what might be causing it.

Can you share the logs for more information about the error?
Isaac Tong 0 Reputation points

2023-03-25T00:05:47.11+00:00
Unfortunately, the logs are not helpful at all as I continue to keep changing the input for the request_json. It would be great if you have information on how to create the json schema for scoring an image classification model using keras CNN?

e.g. If the model takes a set of columns and values as the test dataset, then the input would be something like

{ "input_data": { "columns": [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22], "index": [0, 1], "data": [ [20000,2,2,1,24,2,2,-1,-1,-2,-2,3913,3102,689,0,0,0,0,689,0,0,0,0], [10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 10, 9, 8] ] } }

What about image classification? I can't send an image through json. what would be the required fields?

Thank you!!!!