Infrared small target detection plays an important role in the infrared search and tracking applications. In recent years, deep learning techniques were introduced to this task and achieved noteworthy effects. Following general object segmentation methods, existing deep learning methods usually processed the image from the global view. However, the imaging locality of small targets and extreme class-imbalance between the target and background pixels were not well-considered by these deep learning methods, which causes the low-efficiency on training and high-dependence on numerous data. A multi-patch attention network (MANet) is proposed in this paper to detect small targets by jointly considering the global and local properties of infrared small target images. From the global view, a supervised attention module trained by the small target spread map is proposed to suppress most background pixels irrelevant with small target features. From the local view, local patches are split from global features and share the same convolution weights with each other in a patch net. By synthesizing the global and local properties, the data-driven framework proposed in this paper has fused multi-scale features for small target detection. Extensive synthetic and real data experiments show that the proposed method achieves the state-of-the-art performance compared with existing both conventional and deep learning methods.