Try to use the more lightweight BiFormer's attention mechanism combined with the UNet network to build a Bi-unet medical segmentation model.

The attention mechanism comes from: 2023 CVPR BiFormer: Vision Transformer with Bi-Level Routing Attention Their code in :https://github.com/rayleizhu/BiFormer paper:Zhu L, Wang X, Ke Z, et al. BiFormer: Vision Transformer with Bi-Level Routing Attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023: 10323-10333.

AND For segmented network architecture, we use the most classic :Swin-unet