Supplementary Material for "Modulo Video Recovery via Selective Spatiotemporal Vision Transformer" (ID: 5522) 

The reconstruction quality on the water surface, road surface, and rooftops is significantly superior to other methods.

The reconstruction results of different methods are similar, with slightly better performance than the baseline methods (UnModNet) in reconstructing the sky (clouds) and rooftops. 

The reconstruction results of different methods are similar, with a slight improvement over the baseline method (UnModNet) in the reconstruction of the wall at the beginning of the video.

The reconstruction results of all methods are suboptimal (with noticeable distortions visible to the naked eye), but SSViT achieves significantly better reconstruction of the sky and the wall compared to other methods.

The reconstruction quality of different methods is similar, with a slight improvement over the baseline method (UnModNet) in recovering the patterns on the windows. 

The reconstruction quality of all methods is suboptimal (with noticeable distortions visible to the naked eye), but SSViT achieves significantly better reconstruction of people and clothing compared to other methods. 

When connecting the bright and dark regions, our method demonstrates superior temporal consistency. 

The reconstruction quality of different methods is similar, but SSViT outperforms the others in recovering water ripples.