CVPR2021

Time Lens: Event-Based Video Frame Interpolation

Stepan Tulyakov, Daniel Gehrig, Stamatios Georgoulis, Julius Erbach, Mathias Gehrig, Yuanyou Li, Davide Scaramuzza

Abstract

Figure 1: Qualitative results comparing our proposed method, Time Lens, with DAIN [3] and BMBC [28]. Our method can interpolate frames in highly-dynamic scenes, such as while spinning an umbrella (top row) and bursting a balloon (bottom row). It does this by combining events (b) and frames (a).