deepface-model

This repository has a Dockerfile for deploying DeepFace Models in Ainize.

Example

requirements

• python3
• deepface==0.0.33 (only for example)

cd example 

python3 example.py --mode verify --model VGGFace --image1 ./sample1.png --image2 ./sample2.png 

mode = [verify, analyze]

model = [VGGFace, OpenFace, FbDeepFace, DeepID, Facenet, Age, Gender, Race, Emotion]

or each branch has a example code.

# in vggface branch

cd example 

python3 example.py 

VGGFace

Usage

You must use detected face image.

# input_shape = (None, 224, 224, 3)
# output_shape = (None, None)
# You can check metadata in https://vggface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/vggface/metadata

URL = "https://vggface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/vggface:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

OpenFace

Usage

You must use detected face image.

# input_shape = (None, 96, 96, 3)
# output_shape = (None, 128)
# You can check metadata in https://openface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/openface/metadata

URL = "https://openface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/openface:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

FbDeepFace

Usage

You must use detected face image.

# input_shape = (None, 152, 152, 3)
# output_shape = (None, 4096)
# You can check metadata in https://fbdeepface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/fbdeepface/metadata

URL = "https://fbdeepface-deepface-model-woomurf.endpoint.ainize.ai/v1/models/fbdeepface:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

DeepID

Usage

You must use detected face image.

# input_shape = (None, 55, 47, 3)
# output_shape = (None, 160)
# You can check metadata in https://deepid-deepface-model-woomurf.endpoint.ainize.ai/v1/models/deepid/metadata

URL = "https://deepid-deepface-model-woomurf.endpoint.ainize.ai/v1/models/deepid:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

Facenet

Usage

You must use detected face image.

# input_shape = (None, 160, 160, 3)
# output_shape = (None, 128)
# You can check metadata in https://facenet-deepface-model-woomurf.endpoint.ainize.ai/v1/models/facenet/metadata

URL = "https://facenet-deepface-model-woomurf.endpoint.ainize.ai/v1/models/facenet:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

Age

Usage

You must use detected face image.

# input_shape = (None, 224, 224, 3)
# output_shape = (None, None)
# You can check metadata in https://age-deepface-model-woomurf.endpoint.ainize.ai/v1/models/age/metadata

URL = "https://age-deepface-model-woomurf.endpoint.ainize.ai/v1/models/age:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

Gender

Usage

You must use detected face image.

# input_shape = (None, 224, 224, 3)
# output_shape = (None, None)
# You can check metadata in https://gender-deepface-model-woomurf.endpoint.ainize.ai/v1/models/gender/metadata

URL = "https://gender-deepface-model-woomurf.endpoint.ainize.ai/v1/models/gender:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

Race

Usage

You must use detected face image.

# input_shape = (None, 224, 224, 3)
# output_shape = (None, None)
# You can check metadata in https://race-deepface-model-woomurf.endpoint.ainize.ai/v1/models/race/metadata

URL = "https://race-deepface-model-woomurf.endpoint.ainize.ai/v1/models/race:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']

Emotion

Usage

You must use detected face image.

# input_shape = (None, 48, 48, 1)
# output_shape = (None, 7)
# You can check metadata in https://emotion-deepface-model-woomurf.endpoint.ainize.ai/v1/models/emotion/metadata

URL = "https://emotion-deepface-model-woomurf.endpoint.ainize.ai/v1/models/emotion:predict"

data = {
    "instances": image 
}

data = json.dumps(data)

response = requests.post(URL, data=data)

response_text = response.text 
result = json.loads(response_text)
result = result['predictions']