Python code to call ML Object Detection API

[Former Member]

Could someone please share an python example of how can we call the Object Detection API? It would help me a lot!

[AndreasForster · Accepted solution]

Hello Renan, Here is some Python code that uses the face feature extractor of MLF. It shows how to create and use the Bearer token and how to pass a file and the header to MLF. So this might give you a start for object detection.

Greetings, Andreas

## Get logon credentials from service key

# Path to service key

service_key_location = "./Service Keys/Servicekey_beta.txt"

file_read = open(service_key_location, "r")

jsonRead = json.loads(file_read.read())

username = jsonRead["clientid"]

password = jsonRead["clientsecret"]

## Get Bearer token

import json

import requests

api_url = 'https://[yoururl].authentication.us10.hana.ondemand.com/oauth/token?grant_type=client_credentials'

auth_values = (username, password)

response = requests.get(api_url, auth=auth_values)

# Extract token from response

json_response = response.json()

api_token = "Bearer " + json_response["access_token"]

print(api_token)

## Identify face on image

# Obtain feature vector of face

nugget_file = "./image.jpg"

api_url = 'https://mlfproduction-face-feature-extractor.cfapps.us10.hana.ondemand.com/api/v2alpha1/image/face-feature-extraction'

files = {'files': open(nugget_file, 'rb')}

headers = {'Authorization': api_token}

response_nugget = requests.post(api_url, files=files, headers=headers)

jsonRead = json.loads(response_nugget.content.decode('utf-8'))

# Extract the detected face and display

bottom = jsonRead['predictions'][0]['faces'][0]['face_location']['bottom']

left = jsonRead['predictions'][0]['faces'][0]['face_location']['left']

right = jsonRead['predictions'][0]['faces'][0]['face_location']['right']

top = jsonRead['predictions'][0]['faces'][0]['face_location']['top']

from PIL import Image

img = Image.open(nugget_file)

area = (left, top, right, bottom)

cropped_img = img.crop(area)

cropped_img.show()

[Former Member]

Thank you for the quick answer, Andreas. You've been very helpful!