MTC-VAE logo MTC-VAE

It can be seen the MTC-VAE correctly reenacts the source video with respect to the driving. Motion can be seen in the shrinking/expand of the object, and the corners of the walls. In some cases, β-TCVAE and dis-VAE fail to preserve object identity (as appearance information seems to be leaked from the driving video). SVG-LP shows a seemingly correct reconstruction, but it can be seen that the resulting video does not contain the motion of the driving video.