AnimeRun: 2D Animation Visual Correspondence from Open Source 3D Movies

TL;DR

AnimeRun introduces a new 2D animation dataset derived from 3D movies, featuring complex scenes and motions to better simulate real anime for evaluating correspondence methods.

Contribution

The paper presents AnimeRun, a novel dataset that captures realistic 2D animation scenes from 3D movies, enhancing the evaluation of correspondence algorithms.

Findings

Existing methods struggle with complex animation motions.

AnimeRun dataset resembles real anime more closely.

Benchmark results reveal limitations of current optical flow and segment matching methods.

Abstract

Existing correspondence datasets for two-dimensional (2D) cartoon suffer from simple frame composition and monotonic movements, making them insufficient to simulate real animations. In this work, we present a new 2D animation visual correspondence dataset, AnimeRun, by converting open source three-dimensional (3D) movies to full scenes in 2D style, including simultaneous moving background and interactions of multiple subjects. Our analyses show that the proposed dataset not only resembles real anime more in image composition, but also possesses richer and more complex motion patterns compared to existing datasets. With this dataset, we establish a comprehensive benchmark by evaluating several existing optical flow and segment matching methods, and analyze shortcomings of these methods on animation data. Data, code and other supplementary materials are available at https://lisiyao21.github.io/projects/AnimeRun.