Contextual information induced attention network for infrared small target detection

Infrared small target detection is a challenging task due to the low contrast, small size and complex background. Despite considerable advancements in accuracy, current methods frequently come with high computational costs. Moreover, single modeling that relies on global features or local details usually cannot take into account the effective fusion of global and local information, which affects the detection accuracy and robustness. To solve the above problems, this paper suggests a contextual information induced attention network (CIANet). The network designs a global semantic interaction module (GSI) to perceive the global contextual information and roughly determine the possible small target area. Then, the local context localization module (LCL) carries out refined processing under the guidance of the coarse localization, reducing the computational complexity while ensuring high accuracy. Finally, fusing feature maps of various scales keeps small target’s position and detail information better. Experimental results show that the proposed method outperforms current methods in terms of efficiency and accuracy.