Two-Stream Convolutional Networks for Dynamic Texture Synthesis

TL;DR

This paper presents a two-stream convolutional network model for dynamic texture synthesis that captures both appearance and motion, enabling the creation of high-quality, novel dynamic textures by matching feature statistics.

Contribution

The paper introduces a novel two-stream ConvNet approach for dynamic texture synthesis that combines appearance and motion features for improved realism and flexibility.

Findings

Generates high-quality dynamic textures matching input appearance and motion.

Enables combining appearance from one texture with motion from another.

Quantitative user study confirms the effectiveness of the synthesis method.

Abstract

We introduce a two-stream model for dynamic texture synthesis. Our model is based on pre-trained convolutional networks (ConvNets) that target two independent tasks: (i) object recognition, and (ii) optical flow prediction. Given an input dynamic texture, statistics of filter responses from the object recognition ConvNet encapsulate the per-frame appearance of the input texture, while statistics of filter responses from the optical flow ConvNet model its dynamics. To generate a novel texture, a randomly initialized input sequence is optimized to match the feature statistics from each stream of an example texture. Inspired by recent work on image style transfer and enabled by the two-stream model, we also apply the synthesis approach to combine the texture appearance from one texture with the dynamics of another to generate entirely novel dynamic textures. We show that our approach generates novel, high quality samples that match both the framewise appearance and temporal evolution of input texture. Finally, we quantitatively evaluate our texture synthesis approach with a thorough user study.