Fig. 4.

An illustrated example of $3\text{D}$ shifted windows. The input size $H^{\prime }\times W^{\prime }\times D^{\prime }$ is $8\times 8\times 8$, and the 3D window size $M\times M\times M$ is $4\times 4\times 4$. As layer l adopts regular window partitioning, the number of windows in layer l is $2\times 2\times 2=8.$ For layer $l+1$, as the windows are shifted by $\left(\frac{S_H}{2},,,\frac{S_W}{2},,,\frac{S_D}{2}\right)=(2,2,2)$ tokens, the number of windows becomes $3\times 3\times 3=27$. Though the number of windows is increased, the efficient batch computation in [19] for the shifted configuration can be followed, such that the final number of windows for computation is still 8