Final and intermediate results generated by our method for VoxCeleb, Tai-Chi-HD, and Robonet

The videos on the left part are the results generated by our method, and the videos on the rightmost column are the results of FOMM (Siarohin et al.). The final generated frame by our method is at the 7th column on the left videos. Images on the left videos (ours) from left to right: 1. source frame 2. source mask 3. driving frame 4. driving mask 5. perturbed driving mask 6. low-res prediction 7. final high-res prediction and 8. refined driving mask.

VoxCeleb

Ours FOMM (Siarohin et al.)

Tai-Chi-HD

RoboNet