AnimeRun: 2D Animation Visual Correspondence from Open Source 3D MoviesDownload PDF

Published: 17 Sept 2022, Last Modified: 04 Jun 2023NeurIPS 2022 Datasets and Benchmarks Readers: Everyone
Keywords: 2D animation, cartoon, correspondence, optical flow, matching
TL;DR: We use open source 3D movies to make a new 2D animation dataset with ground truth optical flow and segment-wise correspondence label.
Abstract: Visual correspondence of 2D animation is the core of many applications and deserves careful study. Existing correspondence datasets for 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 3D movies to full scenes in 2D style, including simultaneous moving background and interactions of multiple subjects. Statistics show that our 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 are available at https://lisiyao21.github.io/projects/AnimeRun.
Supplementary Material: zip
Contribution Process Agreement: Yes
In Person Attendance: Yes
URL: https://lisiyao21.github.io/projects/AnimeRun
Dataset Url: https://lisiyao21.github.io/projects/AnimeRun
License: The dataset is made from open movies of Blender Studio. The source movies are shared within Creative Commons (CC) Attribution License. The dataset is released under CC-BY-NC 4.0 License.
Author Statement: Yes
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 1 code implementation](https://www.catalyzex.com/paper/arxiv:2211.05709/code)
25 Replies

Loading