Extrapolative-Interpolative Cycle-Consistency Learning for Video Frame Extrapolation

TL;DR

This paper introduces a novel cycle-consistency loss combining extrapolation and interpolation modules, improving video frame extrapolation by guiding future frame prediction with intermediate frame predictions.

Contribution

The paper proposes the Extrapolative-Interpolative Cycle loss using a pre-trained interpolation module to enhance extrapolation accuracy in video frame prediction.

Findings

EIC loss improves extrapolation performance on UCF101 and KITTI datasets.

Adding EIC loss to existing models enhances short and long-term frame prediction.

The method guarantees consistent predictions across different future time frames.

Abstract

Video frame extrapolation is a task to predict future frames when the past frames are given. Unlike previous studies that usually have been focused on the design of modules or construction of networks, we propose a novel Extrapolative-Interpolative Cycle (EIC) loss using pre-trained frame interpolation module to improve extrapolation performance. Cycle-consistency loss has been used for stable prediction between two function spaces in many visual tasks. We formulate this cycle-consistency using two mapping functions; frame extrapolation and interpolation. Since it is easier to predict intermediate frames than to predict future frames in terms of the object occlusion and motion uncertainty, interpolation module can give guidance signal effectively for training the extrapolation function. EIC loss can be applied to any existing extrapolation algorithms and guarantee consistent prediction in the short future as well as long future frames. Experimental results show that simply adding EIC loss to the existing baseline increases extrapolation performance on both UCF101 and KITTI datasets.