DIANet: Dense-and-Implicit Attention Network
This work addresses vision problems by introducing a novel attention-sharing framework, though it appears incremental as it builds on existing attention and LSTM methods.
The paper tackles the problem of improving vision tasks by proposing a Dense-and-Implicit Attention (DIA) framework that shares an attention module across network layers to integrate layer-wise information, resulting in significant accuracy improvements on benchmark datasets and strong regularization effects in deep networks.
Attention networks have successfully boosted the performance in various vision problems. Previous works lay emphasis on designing a new attention module and individually plug them into the networks. Our paper proposes a novel-and-simple framework that shares an attention module throughout different network layers to encourage the integration of layer-wise information and this parameter-sharing module is referred as Dense-and-Implicit-Attention (DIA) unit. Many choices of modules can be used in the DIA unit. Since Long Short Term Memory (LSTM) has a capacity of capturing long-distance dependency, we focus on the case when the DIA unit is the modified LSTM (refer as DIA-LSTM). Experiments on benchmark datasets show that the DIA-LSTM unit is capable of emphasizing layer-wise feature interrelation and leads to significant improvement of image classification accuracy. We further empirically show that the DIA-LSTM has a strong regularization ability on stabilizing the training of deep networks by the experiments with the removal of skip connections or Batch Normalization in the whole residual network. The code is released at https://github.com/gbup-group/DIANet.