Face Filter Detection


Our Face Filter Detection API determines whether or not an image contains a face filter, or virtual elements are used to decorate or distort someone’s face. Face filters are popular on social media sites like TikTok, Snapchat, and Instagram and can take a variety of forms. The classes for this model make the distinction between distortive face filters, where the facial features are changed, and non-distortive face filters, where items like hearts or glasses are put over the face.

Detailed Class Descriptions

distorting_face_filter: the image contains a face filter that changes the facial features, such as:

  • Images where part of the face is stretched or shrunk
  • Images with filters that recreate the face in a different style, i.e. animated Disney or anime character

non_distorting_face_filter: the image contains a face filter that does not change the facial features, such as:

  • Images with filters that add glasses, hats, or other accessories to the person without altering their face
  • Images with filters that add emojis or other symbols such as hearts on a face, i.e. on cheeks
  • Images that add minimal makeup to a face without drastically changing any facial features

no_face_filter: the image does not contain any face filter. Includes:

  • Images with text overlay that is fully or partially over someone’s face
  • Images where someone is wearing exaggerated makeup that is not a filter

Request Format

# submit a task with media with url
curl --request POST \
  --url https://api.thehive.ai/api/v2/task/sync \
  --header 'accept: application/json' \
  --header 'authorization: token <API_KEY>' \
  --form 'url=http://hive-public.s3.amazonaws.com/demo_request/faces.jpg'

# submit a task with media with local media file
 curl --request POST \
     --url https://api.thehive.ai/api/v2/task/sync \
     --header 'Authorization: Token <token>' \
     --form 'media=@"<absolute/path/to/file>"'


The JSON output for this model is in the same format as the output for the rest of our classification models. To see an annotated example, you can visit our API reference page.