Véronique Prinet - Motion Selective Prediction for Video Frame Synthesis

Motion Selective Prediction for Video Frame Synthesis

Cat sequence

Experimental protocol: trained on 32 frames, prediction on 30 frames. Context of three frames. Frame size: 105x320.

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

Comparison (B2, Ours-M1, Ours-FDN)