Convolutional Neural Networks (CNNs) has always dominated visual recognition tasks, and it is difficult to link distant information in images due to the size limitation of each convolution filter. Vision Transformer (ViT) can capture features at a distance in an image, but lacks the details of local features. Conformer combines the advantages of both using Convolutional Neural Networks (CNNs) and Attention mechanisms in parallel, but it does not take into account the relationship between different samples. Therefore, we propose a new attention calculation method, Extra-Attention, which can effectively learn intra-sample and inter-sample relationships. In order to combine the advantages of CNNs and Attention, in our work, we proposed a new network Enhanced-Conformer (ENC) based on Conformer, in which the attention mechanism adopts a more efficient computational module Inside And Outside Transformer (IAOT), which contains three parallel Attention: Extra-Attention, Self-Attention, External-Attention. Enhanced-Conformer (ENC) can fuse local features, global features and external features at the same time. We conduct experiments on the commonly used image recognition datasets, the recognition accuracies reach 93.29%, 68.70% and 57.63% on Tiny- ImageNet, CIFAR-10 and CIFAR-100, respectively.