def index():
log.info('Home api are visited')
return '''
def classification(payload):
project = AUTO_ML_PROJECT
location = AUTO_ML_LOCATION
model = payload.get('model_name', AUTO_ML_MODEL)
try:
limit = int(payload.get('limit', 3))
except ValueError as e:
logging.error(e)
raise BadRequest("Unable to process limit value '{}'".format(
payload.get('limit')))
try:
threshold = float(payload.get('threshold', 0.0))
except ValueError as e:
logging.error(e)
raise BadRequest("Unable to process threshold value '{}'".format(
payload.get('threshold')))
try:
weighted = payload.get('weighted', False)
if not isinstance(weighted, bool):
raise TypeError('Weighted is not a boolean')
except TypeError as e:
logging.error(e)
raise BadRequest("Unable to process weighted value '{}'".format(
payload.get('weighted')))
# Create an image object from the image data and validate it
try:
image = payload.get('image')
imagedata = re.sub('^data:image/.+;base64,', '', image)
imagebytes = BytesIO(base64.b64decode(imagedata))
image = Image.open(imagebytes)
except base64.binascii.Error as e:
logging.error(e)
raise BadRequest('Unable to process image data: {}'.format(e))
except TypeError as e:
logging.error(e)
message = e.message.message if isinstance(
e.message, base64.binascii.Error) else e.message
raise BadRequest('Unable to process image data: {}'.format(e.message))
except IOError as e:
logging.error(e)
message = list(e.args)[0]
raise BadRequest(message)
if not isinstance(image, PngImageFile):
raise BadRequest('Only png images are accepted')
try:
prediction_client = automl_v1beta1.PredictionServiceClient()
name = prediction_client.model_path(project, location, AUTO_ML_MODEL)
payload = {'image': {'image_bytes': imagebytes.getvalue()}}
params = {'score_threshold': str(threshold)}
logging.info('Calling AutoML {}'.format(name))
response = prediction_client.predict(name, payload, params)
logging.info(response)
except Exception as e:
logging.error(e)
raise InternalServerError('Error communicating with AutoML')
predictions = []
for prediction in response.payload:
glyph = prediction.display_name
score = prediction.classifiction.score * weights.get(
glyph, 1) if weighted else prediction.classification.score
# Score might have been reduced due to weighting, so need to
# check it again here
if glyph != '--other--' and score >= threshold:
predictions.append({'glyph': glyph, 'score': score})
predictions = sorted(predictions,
key=lambda p: p.get('score'),
reverse=True)
if limit < len(predictions):
predictions = predictions[0:limit]
return jsonify(code=200, success=True, result=predictions)
def classification(payload):
try:
limit = int(payload.get('limit', 3))
except ValueError as e:
logging.error(e)
raise BadRequest("Unable to process limit value '{}'".format(
payload.get('limit')))
try:
threshold = float(payload.get('threshold', 0.0))
except ValueError as e:
logging.error(e)
raise BadRequest("Unable to process threshold value '{}'".format(
payload.get('threshold')))
try:
weighted = payload.get('weighted', False)
if not isinstance(weighted, bool):
raise TypeError('Weighted is not a boolean')
except TypeError as e:
logging.error(e)
raise BadRequest("Unable to process weighted value '{}'".format(
payload.get('weighted')))
# Create an image object from the image data and validate it
try:
print('unpacking image info')
image = payload.get('image')
imagedata = re.sub('^data:image/.+;base64,', '', image)
imagebytes = BytesIO(base64.b64decode(imagedata))
image = Image.open(imagebytes)
except base64.binascii.Error as e:
logging.error(e)
raise BadRequest('Unable to process image data: {}'.format(e))
except TypeError as e:
logging.error(e)
message = e.message.message if isinstance(
e.message, base64.binascii.Error) else e.message
raise BadRequest('Unable to process image data: {}'.format(e.message))
except IOError as e:
logging.error(e)
message = list(e.args)[0]
raise BadRequest(message)
if not isinstance(image, PngImageFile):
raise BadRequest('Only png images are accepted')
try:
client_options = {
'api_endpoint': f'{AI_REGION}-aiplatform.googleapis.com'
}
prediction_client = aiplatform.gapic.PredictionServiceClient(
client_options=client_options)
encoded_content = base64.b64encode(
imagebytes.getvalue()).decode('utf-8')
instance = predict.instance.ImageClassificationPredictionInstance(
content=encoded_content).to_value()
instances = [instance]
parameters = predict.params.ImageClassificationPredictionParams(
confidence_threshold=0.5,
max_predictions=5,
).to_value()
endpoint = prediction_client.endpoint_path(project=AI_PROJECT,
location=AI_REGION,
endpoint=AI_ENDPOINT_ID)
response = prediction_client.predict(endpoint=endpoint,
instances=instances,
parameters=parameters)
logging.info(response)
except Exception as e:
print('Error happened in inference')
logging.error(e)
raise InternalServerError('Error communicating with AutoML')
predictions = []
for prediction in response.predictions:
glyph = prediction['displayNames'][0]
confidence = prediction['confidences'][0]
score = confidence * weights.get(glyph, 1) if weighted else confidence
# Score might have been reduced due to weighting, so need to
# check it again here
if glyph != '--other--' and score >= threshold:
predictions.append({'glyph': glyph, 'score': score})
predictions = sorted(predictions,
key=lambda p: p.get('score'),
reverse=True)
if limit < len(predictions):
predictions = predictions[0:limit]
return jsonify(code=200, success=True, result=predictions)
def index():
log.info('Auth are visited')
return '''
def index():
log.info('Pics are visited')
#user = mongo.db.users.find_one_or_404({"_id": username})
return '''
def index():
log.info('Contact are visited')
return '''