Authors: Xiangcheng Du, Xingjiao Wu, Shufan Wu, Shuchen Kong, Hao Ye, Yingbin Zheng, Liang He
Description: We extract the feature by the ResNet16. Especially, we use an attention mechanism that can allow the model to focus on the most important segments of features. We use a novel loss to train our network. We use the Stacked bidirectional LTSM network for the prediction.