<p>Semantic segmentation is widely used in fields such as autonomous driving and unmanned aerial vehicle navigation. However, the huge computational burden and redundant parameters limit its application in edge devices such as mobile phones. In this study, we propose an asymmetric lightweight semantic segmentation model via lightweight attention-guided feature enhancement and fusion. Specifically, the proposed model adopts an encoder-decoder structure. In the encoder, we design an asymmetric feature extraction module to extract image information and use the locally sensitive Hash self-attention to enhance the global information. In the decoder, we first use channel attention to filter out the useless information in shallow layers and adopt the spatial attention to refine local features in deep layers. We then fuse the multi-scale features by the gating mechanism. Additionally, we also design an auxiliary loss to supervise the segmentation of small objects. The results on Cityscapes and CamVid show that the proposed model achieves a good balance between accuracy and the number of parameters. It obtains <b>70.68%</b> and <b>72.19%</b> mIoU on the two test datasets with <b>0</b>.<b>86</b>M parameters, respectively. Code is available on <a href="https://github.com/year410/LAANET">https://github.com/year410/LAANET</a></p>