A keras implementation of "Show, Attend and Tell: Neural Image Caption Generation with Visual Attention"