Attention-Based video2text in Torch
This repository holds many Torch implementations of the seq2seq architecture for video captioning. The original paper, titled Sequence to Sequence -- Video to Text, can be found here. The best model uses an attention mechanism before the captioning layer.