Maize is one of the most widely cultivated crops around the world, but the yield of maize is deeply impacted by leaf diseases and the accurate disease recognition of maize is still a challenge in many regions due to the lack of required infrastructures. In this paper, a dual attention deep neural network Res-Atten is proposed to recognize maize leaf diseases. Specifically, to overcome the problem caused by insufficient data, the pre-trained ResNet-50 model is selected as the baseline feature extraction network. To effectively extract features of maize disease areas, the spatial and channel attention models are introduced in the residual blocks to enhance important features and restrain secondary features. Two different datasets are used to verify the proposed Res-Atten network. Compared with other approaches, the proposed network shows more superior performance, the test recognition accuracy achieves 99.35% on the public dataset and 97.96% on the self-collected maize leaf dataset with complex surroundings. Experimental results demonstrate that the effectiveness of the proposed Res-Atten network for maize leaf disease recognition.