docs.rs failed to build dlib-face-recognition-0.1.2
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Face Recognition
Recognising a face via the dlib models provided takes 4 steps:
- First, A face has to be detected in an image. This is done by first converting a [image] to dlibs matrix format, then running it through either of the face detectors.
- Second, face landmarks have to be predicted. This is called prediction because it only really a guess, and no matter what the number of landmarks returned will be the number of landmarks defined in the model. This takes an image and a face rectangle and generates a series of landmark points on the face, nose, mouth, eyes, etc.
- Then the image and these encodings can be run through the face encoding network to generate encodings of the faces. These encodings consist of 128 floating point numbers that represent the face in 128-dimensional space. To determine if two face encodings belong to the same face, the euclideon distance between them can be used. For the dlib encodings, a distance of 0.6 is generally appropriate.