Overview of VeloxSeg. VeloxSeg employs an encoder-decoder architecture with Paired Window Attention (PWA) and Johnson-Lindenstrauss lemma-guided convolution (JLC) on the left, using 1x1 convolution as ...