Home

Awesome

CompreFace Python SDK

CompreFace Python SDK makes face recognition into your application even easier.

Table of content

Requirements

Before using our SDK make sure you have installed CompreFace and Python on your machine.

  1. CompreFace
  2. Python (Version 3.7+)

CompreFace compatibility matrix

CompreFace Python SDK versionCompreFace 0.5.xCompreFace 0.6.x
0.1.0:yellow_circle:
0.6.x:yellow_circle:

Explanation:

Installation

It can be installed through pip:

pip install compreface-sdk

Usage

All these examples you can find in repository inside examples folder.

Initialization

To start using Python SDK you need to import CompreFace object from 'compreface-sdk' dependency.

Then you need to init it with url and port. By default, if you run CompreFace on your local machine, it's http://localhost and 8000 respectively. You can pass optional options object when call method to set default parameters, see reference for more information.

After you initialized CompreFace object you need to init the service object with the api key of your face service. You can use this service object to recognize faces.

However, before recognizing you need first to add faces into the face collection. To do this, get the face collection object from the service object.

from compreface import CompreFace
from compreface.service import RecognitionService
from compreface.collections import FaceCollection
from compreface.collections.face_collections import Subjects

DOMAIN: str = 'http://localhost'
PORT: str = '8000'
API_KEY: str = 'your_face_recognition_key'

compre_face: CompreFace = CompreFace(DOMAIN, PORT)

recognition: RecognitionService = compre_face.init_face_recognition(API_KEY)

face_collection: FaceCollection = recognition.get_face_collection()

subjects: Subjects = recognition.get_subjects()

Adding faces into a face collection

Here is example that shows how to add an image to your face collection from your file system:

image_path: str = 'examples/common/jonathan-petit-unsplash.jpg'
subject: str = 'Jonathan Petit'

face_collection.add(image_path=image_path, subject=subject)

Recognition

This code snippet shows how to recognize unknown face.

image_path: str = 'examples/common/jonathan-petit-unsplash.jpg'

recognition.recognize(image_path=image_path)

Webcam demo

Webcam demo shows how to use CompreFace Recognition and Detection services using Python SDK. In both cases, age, gender and mask plugins are applied.

Follow this link to see the instructions.

Reference

CompreFace Global Object

Global CompreFace Object is used for initializing connection to CompreFace and setting default values for options. Default values will be used in every service method if applicable. If the option’s value is set in the global object and passed as a function argument then the function argument value will be used.

Constructor:

CompreFace(domain, port, options)

ArgumentTypeRequiredNotes
urlstringrequiredURL with protocol where CompreFace is located. E.g. http://localhost
portstringrequiredCompreFace port. E.g. 8000
optionsobjectoptionalDefault values for face recognition services. See more here. AllOptionsDict object can be used in this method

Possible options:

OptionTypeNotes
det_prob_thresholdfloatminimum required confidence that a recognized face is actually a face. Value is between 0.0 and 1.0
limitintegermaximum number of faces on the image to be recognized. It recognizes the biggest faces first. Value of 0 represents no limit. Default value: 0
prediction_countintegermaximum number of subject predictions per face. It returns the most similar subjects. Default value: 1
face_pluginsstringcomma-separated slugs of face plugins. If empty, no additional information is returned. Learn more
statusbooleanif true includes system information like execution_time and plugin_version fields. Default value is false

Example:

from compreface import CompreFace

DOMAIN: str = 'http://localhost'
PORT: str = '8000'

compre_face: CompreFace = CompreFace(domain=DOMAIN, port=PORT, options={
    "limit": 0,
    "det_prob_threshold": 0.8,
    "prediction_count": 1,
    "face_plugins": "calculator,age,gender,landmarks",
    "status": "true"
})

Methods

  1. CompreFace.init_face_recognition(api_key)

Inits face recognition service object.

ArgumentTypeRequiredNotes
api_keystringrequiredFace Recognition Api Key in UUID format

Example:

from compreface.service import RecognitionService

API_KEY: str = 'your_face_recognition_key'

recognition: RecognitionService = compre_face.init_face_recognition(API_KEY)
  1. CompreFace.init_face_detection(api_key)

Inits face detection service object.

ArgumentTypeRequiredNotes
api_keystringrequiredFace Detection Api Key in UUID format

Example:

from compreface.service import DetectionService

DETECTION_API_KEY: str = 'your_face_detection_key'

detection: DetectionService = compre_face.init_face_detection(DETECTION_API_KEY)
  1. CompreFace.init_face_verification(api_key)

Inits face verification service object.

ArgumentTypeRequiredNotes
api_keystringrequiredFace Verification Api Key in UUID format

Example:

from compreface.service import VerificationService

VERIFICATION_API_KEY: str = 'your_face_verification_key'

verify: VerificationService = compre_face.init_face_verification(VERIFICATION_API_KEY)

Options structure

Options is optional field in every request that contains an image. If the option’s value is set in the global object and passed as a function argument then the function argument value will be used.


class DetProbOptionsDict(TypedDict):
    det_prob_threshold: float


class ExpandedOptionsDict(DetProbOptionsDict):
    limit: int
    status: bool
    face_plugins: str


class AllOptionsDict(ExpandedOptionsDict):
    prediction_count: int

OptionTypeNotes
det_prob_thresholdfloatminimum required confidence that a recognized face is actually a face. Value is between 0.0 and 1.0
limitintegermaximum number of faces on the image to be recognized. It recognizes the biggest faces first. Value of 0 represents no limit. Default value: 0
prediction_countintegermaximum number of subject predictions per face. It returns the most similar subjects. Default value: 1
face_pluginsstringcomma-separated slugs of face plugins. If empty, no additional information is returned. Learn more
statusbooleanif true includes system information like execution_time and plugin_version fields. Default value is false

Example of face recognition with object:

recognition.recognize(image_path=image_path, options={
    "limit": 0,
    "det_prob_threshold": 0.8,
    "prediction_count": 1,
    "face_plugins": "calculator,age,gender,landmarks",
    "status": "true"
})

Face Recognition Service

Face recognition service is used for face identification. This means that you first need to upload known faces to face collection and then recognize unknown faces among them. When you upload an unknown face, the service returns the most similar faces to it. Also, face recognition service supports verify endpoint to check if this person from face collection is the correct one. For more information, see CompreFace page.

Recognize Faces from a Given Image

Example

Recognizes all faces from the image. The first argument is the image location, it can be an url, local path or bytes.

recognition.recognize(image_path, options)
ArgumentTypeRequiredNotes
image_pathimagerequiredImage can pass from url, local path or bytes. Max size is 5Mb
optionsobjectoptionalAllOptionsDict object can be used in this method. See more here.

Response:

{
  "result" : [ {
    "age" : {
      "probability": 0.9308982491493225,
      "high": 32,
      "low": 25
    },
    "gender" : {
      "probability": 0.9898611307144165,
      "value": "female"
    },
    "mask" : {
      "probability": 0.9999470710754395,
      "value": "without_mask"
    },
    "embedding" : [ 9.424854069948196E-4, "...", -0.011415496468544006 ],
    "box" : {
      "probability" : 1.0,
      "x_max" : 1420,
      "y_max" : 1368,
      "x_min" : 548,
      "y_min" : 295
    },
    "landmarks" : [ [ 814, 713 ], [ 1104, 829 ], [ 832, 937 ], [ 704, 1030 ], [ 1017, 1133 ] ],
    "subjects" : [ {
      "similarity" : 0.97858,
      "subject" : "subject1"
    } ],
    "execution_time" : {
      "age" : 28.0,
      "gender" : 26.0,
      "detector" : 117.0,
      "calculator" : 45.0,
      "mask": 36.0
    }
  } ],
  "plugins_versions" : {
    "age" : "agegender.AgeDetector",
    "gender" : "agegender.GenderDetector",
    "detector" : "facenet.FaceDetector",
    "calculator" : "facenet.Calculator",
    "mask": "facemask.MaskDetector"
  }
}
ElementTypeDescription
ageobjectdetected age range. Return only if age plugin is enabled
genderobjectdetected gender. Return only if gender plugin is enabled
maskobjectdetected mask. Return only if face mask plugin is enabled.
embeddingarrayface embeddings. Return only if calculator plugin is enabled
boxobjectlist of parameters of the bounding box for this face
probabilityfloatprobability that a found face is actually a face
x_max, y_max, x_min, y_minintegercoordinates of the frame containing the face
landmarksarraylist of the coordinates of the frame containing the face-landmarks.
subjectslistlist of similar subjects with size of <prediction_count> order by similarity
similarityfloatsimilarity that on that image predicted person
subjectstringname of the subject in Face Collection
execution_timeobjectexecution time of all plugins
plugins_versionsobjectcontains information about plugin versions

Get Face Collection

recognition.get_face_collection()

Returns Face collection object

Face collection could be used to manage known faces, e.g. add, list, or delete them.

Face recognition is performed for the saved known faces in face collection, so before using the recognize method you need to save at least one face into the face collection.

More information about face collection and managing examples here

Methods:

Add an Example of a Subject

Example

This creates an example of the subject by saving images. You can add as many images as you want to train the system. Image should contain only one face.

face_collection.add(image_path, subject, options)
ArgumentTypeRequiredNotes
image_pathimagerequiredImage can pass from url, local path or bytes. Max size is 5Mb
subjectstringrequiredis the name you assign to the image you save
optionsobjectoptionalDetProbOptionsDict object can be used in this method. See more here.

Response:

{
  "image_id": "6b135f5b-a365-4522-b1f1-4c9ac2dd0728",
  "subject": "subject1"
}
ElementTypeDescription
image_idUUIDUUID of uploaded image
subjectstringSubject of the saved image

List of All Saved Examples of the Subject

To retrieve a list of subjects saved in a Face Collection:

face_collection.list()

Response:

{
  "faces": [
    {
      "image_id": <image_id>,
      "subject": <subject>
    },
    ...
  ]
}
ElementTypeDescription
image_idUUIDUUID of the face
subjectstring<subject> of the person, whose picture was saved for this api key

Delete All Examples of the Subject by Name

Example

To delete all image examples of the <subject>:

face_collection.delete_all(subject)
ArgumentTypeRequiredNotes
subjectstringoptionalis the name you assign to the image you save. If this parameter is absent, all faces in Face Collection will be removed

Response:

{
    "deleted": <count>
}
ElementTypeDescription
deletedintegerNumber of deleted faces

Delete an Example of the Subject by ID

Example

To delete an image by ID:

face_collection.delete(image_id)
ArgumentTypeRequiredNotes
image_idUUIDrequiredUUID of the removing face

Response:

{
  "image_id": <image_id>,
  "subject": <subject>
}
ElementTypeDescription
image_idUUIDUUID of the removed face
subjectstring<subject> of the person, whose picture was saved for this api key

Verify Faces from a Given Image

Example

face_collection.verify(image_path, image_id, options)

Compares similarities of given image with image from your face collection.

ArgumentTypeRequiredNotes
image_pathimagerequiredImage can pass from url, local path or bytes. Max size is 5Mb
image_idUUIDrequiredUUID of the verifying face
optionsstringObjectExpandedOptionsDict object can be used in this method. See more here.

Response:

{
  "result" : [ {
    "age" : {
      "probability": 0.9308982491493225,
      "high": 32,
      "low": 25
    },
    "gender" : {
      "probability": 0.9898611307144165,
      "value": "female"
    },
    "mask" : {
      "probability": 0.9999470710754395,
      "value": "without_mask"
    },
    "embedding" : [ 9.424854069948196E-4, "...", -0.011415496468544006 ],
    "box" : {
      "probability" : 1.0,
      "x_max" : 1420,
      "y_max" : 1368,
      "x_min" : 548,
      "y_min" : 295
    },
    "landmarks" : [ [ 814, 713 ], [ 1104, 829 ], [ 832, 937 ], [ 704, 1030 ], [ 1017, 1133 ] ],
    "subjects" : [ {
      "similarity" : 0.97858,
      "subject" : "subject1"
    } ],
    "execution_time" : {
      "age" : 28.0,
      "gender" : 26.0,
      "detector" : 117.0,
      "calculator" : 45.0,
      "mask": 36.0
    }
  } ],
  "plugins_versions" : {
    "age" : "agegender.AgeDetector",
    "gender" : "agegender.GenderDetector",
    "detector" : "facenet.FaceDetector",
    "calculator" : "facenet.Calculator",
    "mask": "facemask.MaskDetector"
  }
}
ElementTypeDescription
ageobjectdetected age range. Return only if age plugin is enabled
genderobjectdetected gender. Return only if gender plugin is enabled
maskobjectdetected mask. Return only if face mask plugin is enabled.
embeddingarrayface embeddings. Return only if calculator plugin is enabled
boxobjectlist of parameters of the bounding box for this face
probabilityfloatprobability that a found face is actually a face
x_max, y_max, x_min, y_minintegercoordinates of the frame containing the face
landmarksarraylist of the coordinates of the frame containing the face-landmarks. Return only if landmarks plugin is enabled
similarityfloatsimilarity that on that image predicted person
execution_timeobjectexecution time of all plugins
plugins_versionsobjectcontains information about plugin versions

Get Subjects

recognition.get_subjects()

Returns subjects object

Subjects object allows working with subjects directly (not via subject examples).

More information about subjects here

Methods:

Add a Subject

Example

Create a new subject in Face Collection.

subjects.add(subject)
ArgumentTypeRequiredNotes
subjectstringrequiredis the name of the subject. It can be any string

Response:

{
  "subject": "subject1"
}
ElementTypeDescription
subjectstringis the name of the subject

List Subjects

Example

Returns all subject related to Face Collection.

subjects.list()

Response:

{
  "subjects": [
    "<subject_name1>",
    "<subject_name2>"
  ]
}
ElementTypeDescription
subjectsarraythe list of subjects in Face Collection

Rename a Subject

Example

Rename existing subject. If a new subject name already exists, subjects are merged - all faces from the old subject name are reassigned to the subject with the new name, old subject removed.

subjects.rename(subject, new_name)
ArgumentTypeRequiredNotes
subjectstringrequiredis the name of the subject that will be updated
new_namestringrequiredis the name of the subject. It can be any string

Response:

{
  "updated": "true|false"
}
ElementTypeDescription
updatedbooleanfailed or success

Delete a Subject

Example

Delete existing subject and all saved faces.

subjects.delete(subject)
ArgumentTypeRequiredNotes
subjectstringrequiredis the name of the subject.

Response:

{
  "subject": "subject1"
}
ElementTypeDescription
subjectstringis the name of the subject

Delete All Subjects

Example

Delete all existing subjects and all saved faces.

subjects.delete_all()

Response:

{
  "deleted": "<count>"
}
ElementTypeDescription
deletedintegernumber of deleted subjects

Face Detection Service

Face detection service is used for detecting faces in the image.

Methods:

Detect

Example

detection.detect(image_path, options)

Finds all faces on the image.

ArgumentTypeRequiredNotes
image_pathimagerequiredimage where to detect faces. Image can pass from url, local path or bytes. Max size is 5Mb
optionsstringObjectExpandedOptionsDict object can be used in this method. See more here.

Response:

{
  "result" : [ {
    "age" : {
      "probability": 0.9308982491493225,
      "high": 32,
      "low": 25
    },
    "gender" : {
      "probability": 0.9898611307144165,
      "value": "female"
    },
    "mask" : {
      "probability": 0.9999470710754395,
      "value": "without_mask"
    },
    "embedding" : [ -0.03027934394776821, "...", -0.05117142200469971 ],
    "box" : {
      "probability" : 0.9987509250640869,
      "x_max" : 376,
      "y_max" : 479,
      "x_min" : 68,
      "y_min" : 77
    },
    "landmarks" : [ [ 156, 245 ], [ 277, 253 ], [ 202, 311 ], [ 148, 358 ], [ 274, 365 ] ],
    "execution_time" : {
      "age" : 30.0,
      "gender" : 26.0,
      "detector" : 130.0,
      "calculator" : 49.0,
      "mask": 36.0
    }
  } ],
  "plugins_versions" : {
    "age" : "agegender.AgeDetector",
    "gender" : "agegender.GenderDetector",
    "detector" : "facenet.FaceDetector",
    "calculator" : "facenet.Calculator",
    "mask": "facemask.MaskDetector"
  }
}
ElementTypeDescription
ageobjectdetected age range. Return only if age plugin is enabled
genderobjectdetected gender. Return only if gender plugin is enabled
maskobjectdetected mask. Return only if face mask plugin is enabled.
embeddingarrayface embeddings. Return only if calculator plugin is enabled
boxobjectlist of parameters of the bounding box for this face (on processedImage)
probabilityfloatprobability that a found face is actually a face (on processedImage)
x_max, y_max, x_min, y_minintegercoordinates of the frame containing the face (on processedImage)
landmarksarraylist of the coordinates of the frame containing the face-landmarks. Return only if landmarks plugin is enabled
execution_timeobjectexecution time of all plugins
plugins_versionsobjectcontains information about plugin versions

Face Verification Service

Example

Face verification service is used for comparing two images. A source image should contain only one face which will be compared to all faces on the target image.

Methods:

Verify

verify.verify(source_image_path, target_image_path, options)

Compares two images provided in arguments. Source image should contain only one face, it will be compared to all faces in the target image.

ArgumentTypeRequiredNotes
image_idUUIDrequiredUUID of the verifying face
source_image_pathimagerequiredfile to be verified. Image can pass from url, local path or bytes. Max size is 5Mb
target_image_pathimagerequiredreference file to check the source file. Image can pass from url, local path or bytes. Max size is 5Mb
optionsstringObjectExpandedOptionsDict object can be used in this method. See more here.

Response:

{
  "result" : [{
    "source_image_face" : {
      "age" : {
        "probability": 0.9308982491493225,
        "high": 32,
        "low": 25
      },
      "gender" : {
        "probability": 0.9898611307144165,
        "value": "female"
      },
      "mask" : {
        "probability": 0.9999470710754395,
        "value": "without_mask"
      },
      "embedding" : [ -0.0010271212086081505, "...", -0.008746841922402382 ],
      "box" : {
        "probability" : 0.9997453093528748,
        "x_max" : 205,
        "y_max" : 167,
        "x_min" : 48,
        "y_min" : 0
      },
      "landmarks" : [ [ 92, 44 ], [ 130, 68 ], [ 71, 76 ], [ 60, 104 ], [ 95, 125 ] ],
      "execution_time" : {
        "age" : 85.0,
        "gender" : 51.0,
        "detector" : 67.0,
        "calculator" : 116.0,
        "mask": 36.0
      }
    },
    "face_matches": [
      {
        "age" : {
          "probability": 0.9308982491493225,
          "high": 32,
          "low": 25
        },
        "gender" : {
          "probability": 0.9898611307144165,
          "value": "female"
        },
        "mask" : {
          "probability": 0.9999470710754395,
          "value": "without_mask"
        },
        "embedding" : [ -0.049007344990968704, "...", -0.01753818802535534 ],
        "box" : {
          "probability" : 0.99975,
          "x_max" : 308,
          "y_max" : 180,
          "x_min" : 235,
          "y_min" : 98
        },
        "landmarks" : [ [ 260, 129 ], [ 273, 127 ], [ 258, 136 ], [ 257, 150 ], [ 269, 148 ] ],
        "similarity" : 0.97858,
        "execution_time" : {
          "age" : 59.0,
          "gender" : 30.0,
          "detector" : 177.0,
          "calculator" : 70.0,
          "mask": 36.0
        }
      }],
    "plugins_versions" : {
      "age" : "agegender.AgeDetector",
      "gender" : "agegender.GenderDetector",
      "detector" : "facenet.FaceDetector",
      "calculator" : "facenet.Calculator",
      "mask": "facemask.MaskDetector"
    }
  }]
}
ElementTypeDescription
source_image_faceobjectadditional info about source image face
face_matchesarrayresult of face verification
ageobjectdetected age range. Return only if age plugin is enabled
genderobjectdetected gender. Return only if gender plugin is enabled
maskobjectdetected mask. Return only if face mask plugin is enabled.
embeddingarrayface embeddings. Return only if calculator plugin is enabled
boxobjectlist of parameters of the bounding box for this face
probabilityfloatprobability that a found face is actually a face
x_max, y_max, x_min, y_minintegercoordinates of the frame containing the face
landmarksarraylist of the coordinates of the frame containing the face-landmarks. Return only if landmarks plugin is enabled
similarityfloatsimilarity between this face and the face on the source image
execution_timeobjectexecution time of all plugins
plugins_versionsobjectcontains information about plugin versions

Contributing

Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.

  1. Fork the Project
  2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
  3. Commit your Changes (git commit -m 'Add some AmazingFeature')
  4. Push to the Branch (git push origin feature/AmazingFeature)
  5. Open a Pull Request

After creating your first contributing pull request, you will receive a request to sign our Contributor License Agreement by commenting your pull request with a special message.

Report Bugs

Please report any bugs here.

If you are reporting a bug, please specify:

Submit Feedback

The best way to send us feedback is to file an issue at https://github.com/exadel-inc/compreface-python-sdk/issues.

If you are proposing a feature, please:

License info

CompreFace Python SDK is open-source facial recognition SDK released under the Apache 2.0 license.