Overview

The service allows to perform face detection, landmarks detection (nose, eyes and
lips) and calculate an embedding vector usefull for face tracking and reidentification.

Face and landmarks detection

• Face bounding box represented by 4 float numbers: x and y position of the left top corner, width
  and height of the box. Numbers are normalized relative to the image size.

• Facial landmarks represented by x and y coordinates of 5 keypoints: nose, left and right eyes, left and
  right corners of lips. Numbers are normalized relative to the image size.

Note: Algorithm may detect multiple faces and sets of landmarks in an image.

Embeddings

Embeddings are represented by vector of 512 float values.

Depending on query parameters, embedding vector may be calculated for each face detected in an image or for the whole image
(usefull for cases when an image consists of a single cropped face).

Endpoints

METHOD	URL	DESCRIPTION
GET	https://face-detection14.p.rapidapi.com/v1/version	Get service version.
POST	https://face-detection14.p.rapidapi.com/v1/results	Perform image analysis and get results.

Get version

Returns an actual version of the service in format vX.Y.Z where X is the
version of API.

PROPERTY	DESCRIPTION
Endpoint	https://face-detection14.p.rapidapi.com/v1/version
Method	GET
Query parameters	–
POST parameters	–

Examples

Request:

$ curl -X 'GET' 'https://face-detection14.p.rapidapi.com/v1/version'

Response:

v1.5.0

Analyse image and return results

Performs actual image analysis and responds with results.

PROPERTY	DESCRIPTION
Endpoint	https://face-detection14.p.rapidapi.com/v1/results
Method	POST
Query parameters	detection,embeddings
POST parameters	image, url

Query parameter: detection

The detection query parameter allows client to enable/disable face & landmarks detection.
If client passes True value then the service will perform detection.
Otherwise if client passes False value then the image will be treated as a cropped face-only image and service will skip detection.
In this case landmarks will be skipped in response.

Detection is enabled by default.

Query parameter: embeddings

The embeddings query parameter allows client to enable/disable embeddings calculation.
If client passes True value then the service will perform calculation of embeddings for each face detected in an image.
Otherwise if client passes False value then embeddings will not be calculated.

Embeddings calculation is disabled by default.

Note: If you want to skip face detection and just calculate embeddings for the whole image, use
the following combination of flags: detection=False&embeddings=True.

Response schema

For responses with 200 HTTP code, the type of response is JSON object with the
following schema:

{
  "results": [
    {
      "status": {
        "code": ...,
        "message": ...
      },
      "name": ...,
      "md5": ...,
      "page": ...,
      "width": ...,
      "height": ...,
      "entities": [
        {
          "kind": "objects",
          "name": "face-detector",
          "objects": [
            {
              "entities": [
                {
                  "kind": "vector",
                  "name": "face-embeddings",
                  "vector": ...
                },
                {
                  "kind": "classes",
                  "name": "face",
                  "classes": {
                    "face": ...
                  }
                },
                {
                  "kind": "namedpoints",
                  "name": "face-landmarks",
                  "namedpoints": {
                    "left-eye": [...],
                    "right-eye": [...],
                    "nose-tip": [...],
                    "mouth-left-corner": [...],
                    "mouth-right-corner": [...]
                  }
                }
              ],
              "box": ...
            },
            ...
          ]
        }
      ]
    }
  ]
}

Primary fields:

Name	Type	Description
results[].status.code	string	Status code of image processing: ok or failure.
results[].status.message	string	Human readable explanation for status of image processing.
results[].name	string	Original image name passed in request (e.g. my_image.jpg).
results[].md5	string	MD5 sum of original image passed in request.
results[].page	int	Optinal page number (presented for multipage inputs only).
results[].width	int	Optinal image width (presented for valid inputs only).
results[].height	int	Optinal image height (presented for valid inputs only).
results[].entities[].objects	array	Array of detected faces.
results[].entities[].objects[].box	array	Face bounding box defined by 4 float values.
results[].entities[].objects[].entities[name=face-embeddings].vector	array	Optional embedding vector represented by array with 512 float values.
results[].entities[].objects[].entities[name=face].classes.face	float	Face detection confidence.
results[].entities[].objects[].entities[name=face-landmarks].namedpoints	object	Optional object that represents facial landmarks.

Other fields that are not described above always have the same values.

Passing image

Image can be passed by posting regular “multipart form data” in two alternative ways:

• as binary file in image field
• as URL to some public resource in url field

Image must be a regular JPEG or PNG image (with or without transparency) or PDF file.
Usually such images have extensions: .jpg, .jpeg, .png, .pdf. In case of PDF
each page will be converted to PNG image and processed separately.
The service checks input file by MIME type and accepts the following types:

• image/jpeg
• image/png
• application/pdf

The size of image file must be less than 16Mb.

Examples

Request:

curl -X 'POST' 'https://face-detection14.p.rapidapi.com/v1/results' -F 'image=@faces.jpg'

Response:

{
  "results": [
    {
      "status": {
        "code": "ok",
        "message": "Success"
      },
      "name": "face.jpg",
      "md5": "39cefc732e7e0c2b1b5f6a783dc87141",
      "width": 1024,
      "height": 768,
      "entities": [
        {
          "kind": "objects",
          "name": "face-detector",
          "objects": [
            {
              "box": [
                0.5379649347853466,
                0.2885976606043589,
                0.19544756039146483,
                0.2706363619501567
              ],
              "entities": [
                {
                  "kind": "classes",
                  "name": "face",
                  "classes": {
                    "face": 0.9990811347961426
                  }
                },
                {
                  "kind": "namedpoints",
                  "name": "face-landmarks",
                  "namedpoints": {
                    "left-eye": [
                      0.5917292404174804,
                      0.37885522842407227
                    ],
                    "right-eye": [
                      0.6897146463394165,
                      0.39089276552200314
                    ],
                    "nose-tip": [
                      0.6375282120704651,
                      0.45772279739379884
                    ],
                    "mouth-left-corner": [
                      0.5896168756484985,
                      0.4839122676849365
                    ],
                    "mouth-right-corner": [
                      0.6723905473947525,
                      0.49339886665344235
                    ]
                  }
                }
              ]
            }
          ]
        }
      ]
    }
  ]
}

Possible errors

The service can not process an image

When client sends an image that can not be processed for some reason(s), the service responds with 200 code and returns JSON object in the same format as the format for successful analysis. In this case, the results[].status.code will have failure value and results[].status.message will contain relevant explanation.

Example of possible reasons for the issue:

• Unsupported file MIME type
• Corrupted image
• File passed as URL is too big or non downloadable

Example response for corrupted image:

{
  "results": [
    {
      "status": {
        "code": "failure",
        "message": "Can not load image."
      },
      "name": "file.jpg",
      "md5": "d41d8cd98f00b204e9800998ecf8427e",
      "entities": []
    }
  ]
}

Request size is too big

Request size is limited by approximately 32Mb.
When client sends request that exceeds this limit, the service responds with 413 code.

The typical reason for exceeding this limit is overly large image.
Taking into account additional HTTP overhead, we strongly recommend to not pass image files of size more than 16Mb.

Example response for too big request:

Error: Request Entity Too Large

Your client issued a request that was too large.

Missing image or url

When client sends a request without an image and url, the service responds with 422 code and returns JSON object.

Example response for request without image or url:

{"detail": "Missing image or url field."}