azure-02 cognitive service-custom vision

1. custom vision

Custom Vision Service를 사용하면 사용자 고유의 이미지 분류자를 빌드, 배포 및 개선할 수 있다. 이미지 분류자는 해당 시각적 특성에 따라 이미지에 레이블을 적용하는 AI 서비스이다.

 

2. 학습 준비

!pip install azure-cognitiveservices-vision-customvision

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting azure-cognitiveservices-vision-customvision
  Downloading azure_cognitiveservices_vision_customvision-3.1.0-py2.py3-none-any.whl (62 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 62.0/62.0 kB 3.2 MB/s eta 0:00:00
Collecting azure-common~=1.1
  Downloading azure_common-1.1.28-py2.py3-none-any.whl (14 kB)
Collecting msrest>=0.5.0
  Downloading msrest-0.7.1-py3-none-any.whl (85 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 85.4/85.4 kB 7.4 MB/s eta 0:00:00
Collecting azure-core>=1.24.0
  Downloading azure_core-1.26.4-py3-none-any.whl (173 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 173.9/173.9 kB 13.4 MB/s eta 0:00:00
Collecting isodate>=0.6.0
  Downloading isodate-0.6.1-py2.py3-none-any.whl (41 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.7/41.7 kB 3.9 MB/s eta 0:00:00
Requirement already satisfied: requests-oauthlib>=0.5.0 in /usr/local/lib/python3.9/dist-packages (from msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (1.3.1)
Requirement already satisfied: requests~=2.16 in /usr/local/lib/python3.9/dist-packages (from msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (2.27.1)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.9/dist-packages (from msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (2022.12.7)
Requirement already satisfied: typing-extensions>=4.3.0 in /usr/local/lib/python3.9/dist-packages (from azure-core>=1.24.0->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (4.5.0)
Requirement already satisfied: six>=1.11.0 in /usr/local/lib/python3.9/dist-packages (from azure-core>=1.24.0->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (1.16.0)
Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.9/dist-packages (from requests~=2.16->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (2.0.12)
Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.9/dist-packages (from requests~=2.16->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (3.4)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.9/dist-packages (from requests~=2.16->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (1.26.15)
Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.9/dist-packages (from requests-oauthlib>=0.5.0->msrest>=0.5.0->azure-cognitiveservices-vision-customvision) (3.2.2)
Installing collected packages: azure-common, isodate, azure-core, msrest, azure-cognitiveservices-vision-customvision
Successfully installed azure-cognitiveservices-vision-customvision-3.1.0 azure-common-1.1.28 azure-core-1.26.4 isodate-0.6.1 msrest-0.7.1

custom vision 모듈을 가져온다.

from azure.cognitiveservices.vision.customvision.training import CustomVisionTrainingClient
from azure.cognitiveservices.vision.customvision.training.models import ImageFileCreateEntry
from msrest.authentication import ApiKeyCredentials

필요한 모듈을 가져온다.

endpoint = ''
training_key = ''

endpoint와 trainig_key를 azure에서 복사한다.

publish_iteration_name = 'GreatWallModel'
credentials = ApiKeyCredentials(in_headers={'Training-key': training_key})
trainer = CustomVisionTrainingClient(endpoint, credentials)

이름과 인증을 설정한다.

project_name = 'GreatWall'
project = trainer.create_project(project_name)

프로젝트를 생성한다.

jajangmyeon_tag = trainer.create_tag(project.id, 'jajangmyeon')
jjamppong_tag = trainer.create_tag(project.id, 'jjamppong')
tansuyook_tag = trainer.create_tag(project.id, ' tangsuyook')

각 사진의 태그를 생성한다.

custom vision 사이트에 접속하여 짜장면, 짬뽕, 탕수육 사진을 충분하게 업로드한다.

 

3. 학습 시작

import time

print('Training......')
iteration = trainer.train_project(project.id)
while (iteration.status != 'Completed'):
  iteration = trainer.get_iteration(project.id, iteration.id)
  print('Training status: ' + iteration.status)
  print('Waiting 10 seconds...')
  time.sleep(10)
  
  Training......
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Training
Waiting 10 seconds...
Training status: Completed
Waiting 10 seconds...

학습을 시작한다.

 

4. 새로운 이미지로 테스트

from azure.cognitiveservices.vision.customvision.prediction import CustomVisionPredictionClient

모델을 생성하고 학습했으면 예측하고 추론하는 모듈을 불러온다.

prediction_key = ''
prediction_endpoint = ''

prediction_credential = ApiKeyCredentials(
    in_headers={'Prediction-key': prediction_key})
    
predictor = CustomVisionPredictionClient(prediction_endpoint, 
                                         prediction_credential)

인증에 필요한 prediction_key, prediction_endpoint를 넣고 예측 리소스를 생성한다.

target_image_url = 'https://imagescdn.gettyimagesbank.com/500/201708/jv10947445.jpg'
result = predictor.classify_image_url(project.id, 
                             'Iteration1', 
                             target_image_url)

테스트 이미지가 아닌 새로운 이미지를 넣어준다.

for prediction in result.predictions:
  print('\t' + prediction.tag_name + 
        ': {0:.2f}%'.format(prediction.probability * 100))
        
 jjamppong: 79.11%
 tangsuyook: 27.16%
 jajangmyeon: 0.11%

결과를 확인한다.