Probably don't need to create files for detected faces

[blissd]

When Fotema detects a face it creates two files, one PNG file that is an extracted image of the exact reported bounds of the detected face, and a second PNG file centred on spot between the eyes and is more suitable for a thumbnail.

The PNG extracted for the detected face bounds is then used as an input for the face recognition processing phase to find similar faces. However... I don't think Fotema really needs to extract the detected face to a separate file to make this work. Instead we just need the bounds (which is saved to Fotema's database) and the source image, for which Fotema also already knows.

I think we could both reduce face detection processing time and file storage costs by no longer extracting the face at detection time and instead just using the face detection bounds and original source image for the inputs for face recognition.