Davis E. King

I'm not going to do it. But maybe someone will contribute :)

The python bindings are in tools/python/src and are relatively easy to add to. I'm busy with other things, so I'm probably not going to do this any time soon. But...

Thanks :) You should be able to implement CTC loss by writing a new loss class that implements [this interface](http://dlib.net/dlib/dnn/loss_abstract.h.html#EXAMPLE_LOSS_LAYER_). Adding something like BLSTM is more complicated since the tooling...

There needs to be a straightforward way for weight sharing and really also some kind of way to specify graphs with cycles rather than explicitly unrolling everything into a DAG.

I'm not planning on it, but maybe someone will do it and submit a PR :)

Sounds cool :)

You can use the trillion faces or ms celeb datasets. They are already available and very large.

No

This example program goes into the details: http://dlib.net/dnn_metric_learning_on_images_ex.cpp.html

The notes on how this model was trained are in the example program that shows how to use it: http://dlib.net/dnn_face_recognition_ex.cpp.html