Added

• original version, probably closest to the one from Kevin Zakka
• but in Tensorflow (Eager) instead of PyTorch