Caffe

CAFFE (Convolutional Architecture for Fast Feature Embedding) is a deep learning framework, originally developed at UC Berkeley.

It is open source, under a BSD license. It is written in C++, with a Python interface.