Comparing Correspondences: Video Prediction with Correspondence-wise Losses

TL;DR

This paper introduces a simple modification to image similarity metrics for video prediction, using optical flow to improve the robustness to positional errors, resulting in clearer and more accurate predictions without changing the underlying networks.

Contribution

It proposes a novel correspondence-wise loss based on optical flow for video prediction, enhancing prediction quality without altering existing network architectures.

Findings

Improved visual sharpness and accuracy in video predictions.

Effective across various video prediction and interpolation tasks.

Achieved strong performance with simple network architectures.

Abstract

Image prediction methods often struggle on tasks that require changing the positions of objects, such as video prediction, producing blurry images that average over the many positions that objects might occupy. In this paper, we propose a simple change to existing image similarity metrics that makes them more robust to positional errors: we match the images using optical flow, then measure the visual similarity of corresponding pixels. This change leads to crisper and more perceptually accurate predictions, and does not require modifications to the image prediction network. We apply our method to a variety of video prediction tasks, where it obtains strong performance with simple network architectures, and to the closely related task of video interpolation. Code and results are available at our webpage: https://dangeng.github.io/CorrWiseLosses