This is my bachelor thesis. I build a attention-based Unet model to do segmentation. Particularly, Unet has three components: Encoder, Bridge, and Decoder. I replace usual CNN blocks in components by attention-based blocks likes: Lambda-block or MHSA block.