def test_send_gender_feedback(self):
""" send face feedback """
feedback_info = FeedbackInfo(end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get(model_id=GENERAL_MODEL_ID)
identities = [
Concept(concept_id='xx', value=True),
Concept(concept_id='x2', value=False)
]
genders = [
Concept(concept_id='male', value=True),
Concept(concept_id='female', value=False)
]
face = Face(identity=FaceIdentity(identities),
gender_appearance=FaceGenderAppearance(genders))
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8)), face=face)
]
ret = m.send_region_feedback(input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_send_multicultural_feedback(self):
""" send face feedback """
feedback_info = FeedbackInfo(end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get('general-v1.3')
identities = [
Concept(concept_id='xx', value=True),
Concept(concept_id='x2', value=False)
]
cultures = [
Concept(concept_id='american', value=True),
Concept(concept_id='asian', value=False)
]
face = Face(
identity=FaceIdentity(identities),
multicultural_appearance=FaceMulticulturalAppearance(cultures))
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8)), face=face)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_concept_object(self):
""" test concept object """
con = Concept("my_concept_1")
self._verify_concept(con)
data = con.dict()
self.assertIn('name', data)
con = Concept("my_concept_1", concept_id='id2', value=0)
self._verify_concept(con)
data = con.dict()
self.assertIn('name', data)
self.assertIn('id', data)
self.assertIn('value', data)
def test_send_complete_multicultural_feedback(self):
""" send face feedback """
feedback_info = FeedbackInfo(
end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get('general-v1.3')
identities = [Concept(concept_id='xx', value=True), Concept(concept_id='x2', value=False)]
ages = [Concept(concept_id='1', value=True), Concept(concept_id='2', value=False)]
genders = [Concept(concept_id='male', value=True), Concept(concept_id='female', value=False)]
cultures = [
Concept(concept_id='american', value=True),
Concept(concept_id='asian', value=False)
]
face = Face(
identity=FaceIdentity(identities),
age_appearance=FaceAgeAppearance(ages),
gender_appearance=FaceGenderAppearance(genders),
multicultural_appearance=FaceMulticulturalAppearance(cultures))
regions = [Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8)), face=face)]
ret = m.send_region_feedback(
input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_send_region_feedback(self):
""" send region feedback with concepts """
feedback_info = FeedbackInfo(end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get('general-v1.3')
concepts = [Concept(concept_id='ab', value=False)]
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8),
feedback_type='accurate'),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
concepts = [
Concept(concept_id='ab', value=False),
Concept(concept_id='bb', value=True)
]
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8)),
concepts=concepts),
Region(RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.8, 0.85)),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_predict_image_url_select_concepts(self):
# url, with select_concepts, by name
m = self.app.models.get('general-v1.3')
select_concepts = [
Concept(concept_name='beer'),
Concept(concept_name='vehicle')
]
res = m.predict_by_url(urls[0], select_concepts=select_concepts)
self.assertEqual(len(res['outputs'][0]['data']['concepts']), 2)
# url, with select_concepts by id
select_concepts = [
Concept(concept_id='ai_hK1KnTCJ'),
Concept(concept_id='ai_m52MdMR3'),
Concept(concept_id='ai_fN8NZ9JV')
]
res = m.predict_by_url(urls[0], select_concepts=select_concepts)
self.assertEqual(len(res['outputs'][0]['data']['concepts']), 3)
def test_send_region_feedback(self):
""" send region feedback with concepts """
feedback_info = FeedbackInfo(end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get(model_id=GENERAL_MODEL_ID)
concepts = [Concept(concept_id='ab', value=False)]
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8),
feedback_type='accurate'),
concepts=concepts)
]
ret = m.send_region_feedback(input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
concepts = [
Concept(concept_id='ab', value=False),
Concept(concept_id='bb', value=True)
]
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8)),
concepts=concepts),
Region(RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.8, 0.85)),
concepts=concepts)
]
ret = m.send_region_feedback(input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_predict_image_url_select_concepts(self):
# url, with select_concepts, by name
m = self.app.models.get(model_id=GENERAL_MODEL_ID)
select_concepts = [
Concept(concept_name='beer'),
Concept(concept_name='vehicle')
]
res = m.predict_by_url(sample_inputs.METRO_IMAGE_URL,
select_concepts=select_concepts)
self.assertEqual(10000, res['status']['code'])
self.assertEqual(len(res['outputs'][0]['data']['concepts']), 2)
# url, with select_concepts by id
select_concepts = [
Concept(concept_id='ai_hK1KnTCJ'),
Concept(concept_id='ai_m52MdMR3'),
Concept(concept_id='ai_fN8NZ9JV')
]
res = m.predict_by_url(sample_inputs.METRO_IMAGE_URL,
select_concepts=select_concepts)
self.assertEqual(10000, res['status']['code'])
self.assertEqual(len(res['outputs'][0]['data']['concepts']), 3)
def test_concept_object(self):
""" test concept object """
con = Concept("my_concept_1")
self._verify_concept(con)
data = con.dict()
self.assertIn('name', data)
con = Concept("my_concept_1", concept_id='id2', value=0)
self._verify_concept(con)
data = con.dict()
self.assertIn('name', data)
self.assertIn('id', data)
self.assertIn('value', data)
def prediction_generator(predict_model, predict_images, predict_concepts,
min_predict_value):
select_concept_list = [
Concept(concept_id=concept) for concept in predict_concepts
]
try:
model = CL_APP.models.get(predict_model)
for image in predict_images:
# Check whether file need to chunk or not
if image.multiple_chunks:
string_buffer = cStringIO.StringIO()
for chunk in image:
string_buffer.write(chunk)
string_buffer.flush()
img_base64 = base64.encodestring(string_buffer.getvalue())
string_buffer.close()
# Small size file should read direct instead
else:
img_string = image.read()
img_base64 = base64.encodestring(img_string)
if len(select_concept_list) > 0:
predict_result = model.predict_by_base64(
img_base64, select_concepts=select_concept_list)
else:
predict_result = model.predict_by_base64(
img_base64, min_value=min_predict_value)
yield json.dumps(predict_result)
except ApiError as api_err:
yield {'code': 500, 'message': api_err.error_details}
except ApiClientError:
yield {'code': 500, 'message': 'Error while predict image'}
def clarifai_detection_object_in_image(base64_image):
start = time.time()
select_concept_list = list()
for predefined_concept in predefined_concepts:
select_concept_list.append(Concept(concept_name=predefined_concept))
response = model.predict_by_base64(base64_bytes=base64_image,
select_concepts=select_concept_list)
results = []
for clarifai_concept in response['outputs'][0]['data']['concepts']:
print(
str(clarifai_concept['name']) + ': ' +
str(clarifai_concept['value']))
if clarifai_concept['value'] >= 0.1:
results.append(
DetectionResult(tag=clarifai_concept['name'],
score=clarifai_concept['value']).__dict__)
end = time.time()
print('Cl took: ' + str(end - start) + ' with the return results: ' +
str(response))
return results
def test_send_different_region_feedbacks_with_enum_vars(self):
""" send different kind of region feedbacks with enums
- accurate
- misplaced
- not_detected
- false_positive
"""
feedback_info = FeedbackInfo(end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get('general-v1.3')
# positive
concepts = [Concept(concept_id='ab', value=False)]
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8),
feedback_type=FeedbackType.accurate),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# misplaced
regions = [
Region(RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.8, 0.9),
feedback_type=FeedbackType.misplaced),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# misplaced
regions = [
Region(RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.4, 0.5),
feedback_type=FeedbackType.not_detected),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# false_positive
regions = [
Region(RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.4, 0.5),
feedback_type=FeedbackType.false_positive),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url='https://samples.clarifai.com/dog.tiff',
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def test_bulk_update_with_regions(mock_http_client): # type: (mock.Mock) -> None
mock_execute_request = mock_request(
mock_http_client, """
{
"status": {
"code": 10000,
"description": "Ok"
},
"inputs": [
{
"id": "@inputID",
"data": {
"image": {
"url": "@imageURL"
},
"concepts": [
{
"id": "@concept1",
"name": "@concept1",
"value": 1,
"app_id": "@appID"
},
{
"id": "@concept2",
"name": "@concept2",
"value": 0,
"app_id": "@appID"
}
]
},
"created_at": "2019-01-29T15:23:21.188492Z",
"modified_at": "2019-01-29T15:23:21.575667Z",
"status": {
"code": 30200,
"description": "Input image modification success"
}
}
]
}
""")
app = ClarifaiApp()
images = app.inputs.bulk_update(
[
Image(
image_id='@inputID',
regions=[
Region(
region_info=RegionInfo(
bbox=BoundingBox(0.5, 0.5, 1.0, 1.0),
feedback_type=FeedbackType.misplaced),
concepts=[
Concept(concept_id='@concept1', value=True),
Concept(concept_id='@concept2', value=False)
],
region_id='@regionID',
face=Face(
identity=FaceIdentity(concepts=[
Concept(concept_id='@faceConcept1', value=True),
Concept(concept_id='@faceConcept2', value=False),
])))
])
],
action='overwrite')
assert images[0].input_id == '@inputID'
assert images[0].url == '@imageURL'
assert images[0].status.code == 30200
assert_request(
mock_execute_request, 'PATCH', '/v2/inputs', """
{
"inputs": [
{
"id": "@inputID",
"data": {
"regions": [
{
"id": "@regionID",
"region_info": {
"bounding_box": {
"top_row": 0.5,
"left_col": 0.5,
"bottom_row": 1,
"right_col": 1
},
"feedback": "misplaced"
},
"data": {
"concepts": [
{
"id": "@concept1",
"value": 1
},
{
"id": "@concept2",
"value": 0
}
],
"face": {
"identity": {
"concepts": [
{
"id": "@faceConcept1",
"value": 1
},
{
"id": "@faceConcept2",
"value": 0
}
]
}
}
}
}
]
}
}
],
"action": "overwrite"
}
""")
from clarifai.rest import ClarifaiApp
from clarifai.rest import Image as ClImage
from clarifai.rest import ClarifaiApp, Concept
app = ClarifaiApp(api_key='d4d5faab185e4590b14866212e252f3b')
model = app.models.get('general-v1.3')
select_concept_list = [Concept(concept_name='car')]
print(
model.predict_by_url(
url=
'https://s.driving-tests.org/wp-content/uploads/2012/02/back-parking.jpg',
select_concepts=select_concept_list))
def test_send_different_region_feedbacks(self):
""" send different kind of region feedbacks
- accurate
- misplaced
- not_detected
- false_positive
"""
feedback_info = FeedbackInfo(
end_user_id='robert_python_test_key',
session_id='from_your_browser',
event_type='annotation',
output_id='oooooooid')
m = self.app.models.get('general-v1.3')
# positive
concepts = [Concept(concept_id='ab', value=False)]
regions = [
Region(
RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.7, 0.8), feedback_type='accurate'),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# misplaced
regions = [
Region(
RegionInfo(bbox=BoundingBox(0.3, 0.2, 0.8, 0.9), feedback_type='misplaced'),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# misplaced
regions = [
Region(
RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.4, 0.5), feedback_type='not_detected'),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
# false_positive
regions = [
Region(
RegionInfo(bbox=BoundingBox(0.1, 0.2, 0.4, 0.5), feedback_type='false_positive'),
concepts=concepts)
]
ret = m.send_region_feedback(
input_id='xyz',
url=sample_inputs.DOG_TIFF_IMAGE_URL,
concepts=['matid2'],
not_concepts=['lambo'],
regions=regions,
feedback_info=feedback_info)
self.assertEqual(ret['status']['code'], 10000)
def make_prediction(request):
contexts = dict()
if request.method == 'POST':
predict_model = request.POST.get('predict_model', '')
predict_concepts = request.POST.get('predict_concepts', '')
predict_concepts = [lambda: predict_concepts.split(','),
lambda: []][predict_concepts == '']()
select_concept_list = [
Concept(concept_id=concept) for concept in predict_concepts
]
min_predict_value = request.POST.get('min_predict_value', '0')
min_predict_value = float(min_predict_value)
predict_images = request.FILES.getlist('predict_images')
if not predict_model == '':
if None not in predict_images:
return StreamingHttpResponse(gen_test(),
content_type='text/plain')
# return StreamingHttpResponse(prediction_generator(predict_model=predict_model,
# predict_images=predict_images,
# predict_concepts=predict_concepts,
# min_predict_value=min_predict_value),
# content_type='application/json')
# try:
# model = CL_APP.models.get(predict_model)
# prediction_data = list()
# for image in predict_images:
# # Check whether file need to chunk or not
# if image.multiple_chunks:
# string_buffer = cStringIO.StringIO()
# for chunk in image:
# string_buffer.write(chunk)
# string_buffer.flush()
# img_base64 = base64.encodestring(string_buffer.getvalue())
# string_buffer.close()
# # Small size file should read direct instead
# else:
# img_string = image.read()
# img_base64 = base64.encodestring(img_string)
# if len(select_concept_list) > 0:
# predict_result = model.predict_by_base64(img_base64,
# select_concepts=select_concept_list)
# else:
# predict_result = model.predict_by_base64(img_base64, min_value=min_predict_value)
# prediction_data.append(predict_result)
# contexts.update({
# 'code': 10000,
# 'message': 'Successfully made prediction of '+str(len(predict_images))+' images.',
# 'prediction_data': prediction_data
# })
# except ApiError as api_err:
# contexts.update({
# 'code': 500,
# 'message': api_err.error_details
# })
# pass
# except ApiClientError:
# pass
else:
status_alert = {
'code': 400,
'title': 'Failed',
'message': 'Images must be presented in request form',
'type': 'error'
}
contexts.update({'status_alert': status_alert})
else:
contexts.update({'code': 300, 'message': 'Model must be specify'})
else:
contexts.update({
'code':
405,
'message':
'Method not allowed or Request data not valid'
})
return JsonResponse(contexts)
