I've created and trained an image classification model using Azure Custom Vision (Cognitive Services) and published the model with API.
Now, I've written a simple code in Python which takes an image from given URL and calls the API. However, I'm still getting this error even though the image surely exists:
with open(URL, "rb") as image_contents: FileNotFoundError: [Errno 2]
No such file or directory:
'https://upload.wikimedia.org/wikipedia/commons/5/55/Dalailama1_20121014_4639.jpg'
The code is as below:
from azure.cognitiveservices.vision.customvision.prediction import CustomVisionPredictionClient
ENDPOINT = "https://westeurope.api.cognitive.microsoft.com/"
PROJECT_ID = "bbed3f99-4199-4a17-81f2-df83f0659be3"
# Replace with a valid key
prediction_key = "<my prediction key>"
prediction_resource_id = "/subscriptions/97c4e143-9c0c-4f1e-b880-15492e327dd1/resourceGroups/WestEurope/providers/Microsoft.CognitiveServices/accounts/HappyAI"
publish_iteration_name = "Iteration5"
# Classify image
URL = "https://upload.wikimedia.org/wikipedia/commons/5/55/Dalailama1_20121014_4639.jpg"
# Now there is a trained endpoint that can be used to make a prediction
predictor = CustomVisionPredictionClient(prediction_key, endpoint=ENDPOINT)
with open(URL, "rb") as image_contents:
results = predictor.classify_image(
PROJECT_ID, publish_iteration_name, image_contents.read())
# Display the results.
for prediction in results.predictions:
print("\t" + prediction.tag_name +
": {0:.2f}%".format(prediction.probability * 100))
Help would be appreciated!
Thanks in advance!
There are two ways to give an image to the Cognitive Service. You are mixing both ;)
1) Provide a URL to an image that is accessible over the internet. You do this by sending a JSON to the service:
{"url":"https://sample.com/myimage.png"}
2) Upload the image as binary in the POST request.
Source: https://learn.microsoft.com/en-us/azure/cognitive-services/custom-vision-service/use-prediction-api#get-the-url-and-prediction-key
Your issue is that you are trying to use open() for method 2. However, this does not work with remote files in Python. If you want to do this (instead of method 1), use for example urllib2.urlopen like this.
Related
I want to deploy my trained tensorflow model to the amazon sagemaker, I am following the official guide here: https://aws.amazon.com/blogs/machine-learning/deploy-trained-keras-or-tensorflow-models-using-amazon-sagemaker/ to deploy my model using jupyter notebook.
But when I try to use code:
predictor = sagemaker_model.deploy(initial_instance_count=1, instance_type='ml.t2.medium')
It gives me the following error message:
ValueError: Error hosting endpoint sagemaker-tensorflow-2019-08-07-22-57-59-547: Failed Reason: The image '520713654638.dkr.ecr.us-west-1.amazonaws.com/sagemaker-tensorflow:1.12-cpu-py3 ' does not exist.
I think the tutorial does not tell me to create an image, and I do not know what to do.
import boto3, re
from sagemaker import get_execution_role
role = get_execution_role()
# make a tar ball of the model data files
import tarfile
with tarfile.open('model.tar.gz', mode='w:gz') as archive:
archive.add('export', recursive=True)
# create a new s3 bucket and upload the tarball to it
import sagemaker
sagemaker_session = sagemaker.Session()
inputs = sagemaker_session.upload_data(path='model.tar.gz', key_prefix='model')
from sagemaker.tensorflow.model import TensorFlowModel
sagemaker_model = TensorFlowModel(model_data = 's3://' + sagemaker_session.default_bucket() + '/model/model.tar.gz',
role = role,
framework_version = '1.12',
entry_point = 'train.py',
py_version='py3')
%%time
#here I fail to deploy the model and get the error message
predictor = sagemaker_model.deploy(initial_instance_count=1,
instance_type='ml.m4.xlarge')
https://github.com/aws/sagemaker-python-sdk/issues/912#issuecomment-510226311
As mentioned in the issue
Python 3 isn't supported using the TensorFlowModel object, as the container uses the TensorFlow serving api library in conjunction with the GRPC client to handle making inferences, however the TensorFlow serving api isn't supported in Python 3 officially, so there are only Python 2 versions of the containers when using the TensorFlowModel object.
If you need Python 3 then you will need to use the Model object defined in #2 above. The inference script format will change if you need to handle pre and post processing. https://github.com/aws/sagemaker-tensorflow-serving-container#prepost-processing.
I'm working for the first time with IBM Watson Visual Recognition. My Python app needs to pass images that it's managing in memory to the service. However, the rather limited documentation and sample code I've been able to find from IBM shows calls to the API as referencing saved files. The file is passed to the call as an io.BufferedReader.
with open(car_path, 'rb') as images_file:
car_results = service.classify(
images_file=images_file,
threshold='0.1',
classifier_ids=['default']
).get_result()
My application will be working with images from memory and I don't want to have to save every image to file before I can make a call. I tried replacing the BufferedReader with an io.BytesIO stream, and I got back an error saying I was missing an images_filename param. When I added a mock filename (e.g. 'xyz123.jpg') I get back the following error:
TypeError: a bytes-like object is required, not 'float'
Can I make calls to the analysis API using an image from memory? If so, how?
EDIT:
This is essentially what I'm trying to do:
def analyze_image(pillow_img: PIL.Image):
byte_stream = io.BytesIO()
pillow_img.save(byte_stream, format='JPEG')
bytes_img = byte_stream.getvalue()
watson_vr = VisualRecognitionV3(
'2019-04-30',
url='https://gateway.watsonplatform.net/visual-recognition/api',
iam_apikey='<API KEY>'
)
result_json = watson_vr.classify(
images_file=bytes_img,
threshold=0.1,
classifier_ids=['default']
).get_result()
Thanks
How about
bytes_img = byte_stream.getbuffer()
...
result_json = watson_vr.classify(
images_file=bytes_img,
threshold=0.1,
classifier_ids=['default']
).get_result()
or
with byte_stream as images_file:
result_json = watson_vr.classify(
images_file=images_file,
threshold='0.1',
classifier_ids=['default']
).get_result()
I was trying to create a script that feeds articles through the classification tool of the Natural Language API and I found a tutorial that does exactly that. I was following this simple tutorial to get an intro into Google Cloud and the Natural Language API.
The end result is supposed to be a script that sends a bunch of new articles from the Google Cloud Storage to the Natural Language API to classify the articles and then save the whole thing into a table created in BigQuery.
I was following the example fine, but when running the final script I get the following error:
Traceback (most recent call last):
File "classify-text.py", line 39, in <module>
errors = bq_client.create_rows(table, rows_for_bq)
AttributeError: 'Client' object has no attribute 'create_rows'
The full script is:
from google.cloud import storage, language, bigquery
# Set up our GCS, NL, and BigQuery clients
storage_client = storage.Client()
nl_client = language.LanguageServiceClient()
# TODO: replace YOUR_PROJECT with your project name below
bq_client = bigquery.Client(project='Your_Project')
dataset_ref = bq_client.dataset('news_classification')
dataset = bigquery.Dataset(dataset_ref)
table_ref = dataset.table('article_data')
table = bq_client.get_table(table_ref)
# Send article text to the NL API's classifyText method
def classify_text(article):
response = nl_client.classify_text(
document=language.types.Document(
content=article,
type=language.enums.Document.Type.PLAIN_TEXT
)
)
return response
rows_for_bq = []
files = storage_client.bucket('text-classification-codelab').list_blobs()
print("Got article files from GCS, sending them to the NL API (this will take ~2 minutes)...")
# Send files to the NL API and save the result to send to BigQuery
for file in files:
if file.name.endswith('txt'):
article_text = file.download_as_string()
nl_response = classify_text(article_text)
if len(nl_response.categories) > 0:
rows_for_bq.append((article_text, nl_response.categories[0].name, nl_response.categories[0].confidence))
print("Writing NL API article data to BigQuery...")
# Write article text + category data to BQ
errors = bq_client.create_rows(table, rows_for_bq)
assert errors == []
You are using deprecated methods; these methods were marked as obsolete in version 0.29, and removed altogether in version 1.0.0.
You should use client.insert_rows() instead; the method accepts the same arguments:
errors = bq_client.insert_rows(table, rows_for_bq)
I have searched the web far and wide for a still working example of uploading a photo to facebook through the Python API (Python for Facebook). Questions like this have been asked on stackoverflow before but non of the answers I have found work anymore.
What I got working is:
import facebook as fb
cfg = {
"page_id" : "my_page_id",
"access_token" : "my_access_token"
}
api = get_api(cfg)
msg = "Hello world!"
status = api.put_wall_post(msg)
where I have defined the get_api(cfg) function as this
graph = fb.GraphAPI(cfg['access_token'], version='2.2')
# Get page token to post as the page. You can skip
# the following if you want to post as yourself.
resp = graph.get_object('me/accounts')
page_access_token = None
for page in resp['data']:
if page['id'] == cfg['page_id']:
page_access_token = page['access_token']
graph = fb.GraphAPI(page_access_token)
return graph
And this does indeed post a message to my page.
However, if I instead want to upload an image everything goes wrong.
# Upload a profile photo for a Page.
api.put_photo(image=open("path_to/my_image.jpg",'rb').read(), message='Here's my image')
I get the dreaded GraphAPIError: (#324) Requires upload file for which non of the solutions on stackoverflow works for me.
If I instead issue the following command
api.put_photo(image=open("path_to/my_image.jpg",'rb').read(), album_path=cfg['page_id'] + "/picture")
I get GraphAPIError: (#1) Could not fetch picture for which I haven't been able to find a solution either.
Could someone out there please point me in the right direction of provide me with a currently working example? It would be greatly appreciated, thanks !
A 324 Facebook error can result from a few things depending on how the photo upload call was made
a missing image
an image not recognised by Facebook
incorrect directory path reference
A raw cURL call looks like
curl -F 'source=#my_image.jpg' 'https://graph.facebook.com/me/photos?access_token=YOUR_TOKEN'
As long as the above calls works, you can be sure the photo agrees with Facebook servers.
An example of how a 324 error can occur
touch meow.jpg
curl -F 'source=#meow.jpg' 'https://graph.facebook.com/me/photos?access_token=YOUR_TOKEN'
This can also occur for corrupted image files as you have seen.
Using .read() will dump the actual data
Empty File
>>> image=open("meow.jpg",'rb').read()
>>> image
''
Image File
>>> image=open("how.png",'rb').read()
>>> image
'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00...
Both of these will not work with the call api.put_photo as you have seen and Klaus D. mentioned the call should be without read()
So this call
api.put_photo(image=open("path_to/my_image.jpg",'rb').read(), message='Here's my image')
actually becomes
api.put_photo('\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00...', message='Here's my image')
Which is just a string, which isn't what is wanted.
One needs the image reference <open file 'how.png', mode 'rb' at 0x1085b2390>
I know this is old and doesn't answer the question with the specified API, however, I came upon this via a search and hopefully my solution will help travelers on a similar path.
Using requests and tempfile
A quick example of how I do it using the tempfile and requests modules.
Download an image and upload to Facebook
The script below should grab an image from a given url, save it to a file within a temporary directory and automatically cleanup after finished.
In addition, I can confirm this works running on a Flask service on Google Cloud Run. That comes with the container runtime contract so that we can store the file in-memory.
import tempfile
import requests
# setup stuff - certainly change this
filename = "your-desired-filename"
filepath = f"{directory}/{filename}"
image_url = "your-image-url"
act_id = "your account id"
access_token = "your access token"
# create the temporary directory
temp_dir = tempfile.TemporaryDirectory()
directory = temp_dir.name
# stream the image bytes
res = requests.get(image_url, stream=True)
# write them to your filename at your temporary directory
# assuming this works
# add logic for non 200 status codes
with open(filepath, "wb+") as f:
f.write(res.content)
# prep the payload for the facebook call
files = {
"filename": open(filepath, "rb"),
}
url = f"https://graph.facebook.com/v10.0/{act_id}/adimages?access_token={access_token}"
# send the POST request
res = requests.post(url, files=files)
res.raise_for_status()
if res.status_code == 200:
# get your image data back
image_upload_data = res.json()
temp_dir.cleanup()
if "images" in image_upload_data:
return image_upload_data["images"][filepath.split("/")[-1]]
return image_upload_data
temp_dir.cleanup() # paranoid: just in case an error isn't raised
I am trying to use this api: ImageSimilarity with the Python endpoint.
Any ideas on how to load local images in the API? The example shows:
[ "data://zskurultay/ImageDemo/butterfly1.png",
"data://zskurultay/ImageDemo/butterfly1.png"]
Yet when I pass as input1 a local image for example with:data://home/username/path/to/image.png yelds the error:
Exception: algorithmia.api.APIException: Unexpected API response, status 400,
url http://172.17.42.1:4160/v1/data/home/username/path/to/image.png:
{"error":{"message":"Path invalid"}}
data:// is for Algorithmia cloud images. (You can create your own bucket in their cloud.) To access my local images I try to use different code like,
algo = client.algo( '... alg name ....')
image = base64.b64encode( open( infile, "rb").read())
image_alg = algo.pipe( {'image':'data:image/jpg;base64,' + image.decode('ascii')})
return image_alg.result
But I am still searching for the right tools.