Understanding Object Dynamics for Interactive Image-to-Video Synthesis

TL;DR

This paper introduces a generative model that predicts natural object deformations in response to local pixel-level manipulations, enabling interactive control over object dynamics in images without requiring physical scene data.

Contribution

It presents a novel approach that learns object dynamics from videos to enable local interactive deformation prediction on static images, applicable to unseen objects.

Findings

Effective in predicting realistic object deformations

Outperforms common video prediction frameworks

Works on diverse object categories

Abstract

What would be the effect of locally poking a static scene? We present an approach that learns naturally-looking global articulations caused by a local manipulation at a pixel level. Training requires only videos of moving objects but no information of the underlying manipulation of the physical scene. Our generative model learns to infer natural object dynamics as a response to user interaction and learns about the interrelations between different object body regions. Given a static image of an object and a local poking of a pixel, the approach then predicts how the object would deform over time. In contrast to existing work on video prediction, we do not synthesize arbitrary realistic videos but enable local interactive control of the deformation. Our model is not restricted to particular object categories and can transfer dynamics onto novel unseen object instances. Extensive experiments on diverse objects demonstrate the effectiveness of our approach compared to common video prediction frameworks. Project page is available at https://bit.ly/3cxfA2L .