зеркало из https://github.com/Azure/azcv.git
77 строки
1.9 KiB
Python
77 строки
1.9 KiB
Python
# -*- coding: utf-8 -*-
|
|
|
|
# Copyright (c) Microsoft Corporation. All rights reserved.
|
|
# Licensed under the MIT License.
|
|
# Author: Graham.Williams@togaware.com
|
|
#
|
|
# A script to identify a landmark in a photo.
|
|
#
|
|
# ml landmark azcv <path>
|
|
#
|
|
|
|
from azure.cognitiveservices.vision.computervision import ComputerVisionClient
|
|
from msrest.authentication import CognitiveServicesCredentials
|
|
|
|
import os
|
|
import sys
|
|
#import time
|
|
import argparse
|
|
|
|
from mlhub.pkg import azkey, is_url
|
|
from mlhub.utils import get_cmd_cwd
|
|
|
|
# ----------------------------------------------------------------------
|
|
# Parse command line arguments
|
|
# ----------------------------------------------------------------------
|
|
|
|
option_parser = argparse.ArgumentParser(add_help=False)
|
|
|
|
option_parser.add_argument(
|
|
'path',
|
|
help='path or url to image')
|
|
|
|
args = option_parser.parse_args()
|
|
|
|
# ----------------------------------------------------------------------
|
|
|
|
SERVICE = "Computer Vision"
|
|
KEY_FILE = os.path.join(os.getcwd(), "private.txt")
|
|
|
|
# Request subscription key and endpoint from user.
|
|
|
|
subscription_key, endpoint = azkey(KEY_FILE, SERVICE, verbose=False)
|
|
|
|
# Set credentials.
|
|
|
|
credentials = CognitiveServicesCredentials(subscription_key)
|
|
|
|
# Create client.
|
|
|
|
client = ComputerVisionClient(endpoint, credentials)
|
|
|
|
# Check the URL supplied. Also want to support local file.
|
|
|
|
# Send image to azure to identify landmark
|
|
|
|
# url = "https://images.pexels.com/photos/338515/pexels-photo-338515.jpeg"
|
|
|
|
url = args.path
|
|
|
|
domain = "landmarks"
|
|
language = "en"
|
|
|
|
if is_url(url):
|
|
analysis = client.analyze_image_by_domain(domain, url, language)
|
|
else:
|
|
path = os.path.join(get_cmd_cwd(), url)
|
|
with open(path, 'rb') as fstream:
|
|
analysis = client.analyze_image_by_domain_in_stream(domain, fstream, language)
|
|
|
|
for landmark in analysis.result["landmarks"]:
|
|
print('{},{}'.format(round(landmark["confidence"],2), landmark["name"], ))
|
|
|
|
# Write results to stdout
|
|
|
|
# Generate locally annotated image
|
|
|