Véronique Prinet - Motion Selective Prediction for Video Frame Synthesis

Motion Selective Prediction for Video Frame Synthesis

Garden sequence


Experimental protocol: trained on 30 frames, prediction on 23 frames. Context of three frames. Frame size: 100x300.


Training (12fps) Prediction - Ours-FDN (7fps)

B1 Ours-M1

Comparison (B1, Ours-M1, Ours-FDN)

Background Foreground

Selector activation curves