LinkTo-Anime: A 2D Animation Optical Flow Dataset from 3D Model Rendering
🤖 AI Summary
Existing optical flow datasets lack high-quality benchmarks tailored to cel-animated character motion, hindering progress in animated video understanding and generation. To address this, we introduce CelFlow—the first optical flow dataset specifically designed for hand-drawn anime-style character motion. CelFlow is constructed by driving 3D rigged models from the Mixamo motion library to generate multi-view-consistent renderings; it provides pixel-accurate forward/backward optical flow ground truth, occlusion masks, and synchronized skeletal joint annotations. The dataset comprises 24,230 frames covering characteristic anime motion patterns. We establish a cross-method benchmark, revealing critical performance bottlenecks of mainstream optical flow models on cel animation. Furthermore, leveraging CelFlow significantly improves downstream tasks—including animation generation and line-art coloring—demonstrating its utility in bridging a key gap in visual motion modeling for anime.
Application Category
📝 Abstract
Existing optical flow datasets focus primarily on real-world simulation or synthetic human motion, but few are tailored to Celluloid(cel) anime character motion: a domain with unique visual and motion characteristics. To bridge this gap and facilitate research in optical flow estimation and downstream tasks such as anime video generation and line drawing colorization, we introduce LinkTo-Anime, the first high-quality dataset specifically designed for cel anime character motion generated with 3D model rendering. LinkTo-Anime provides rich annotations including forward and backward optical flow, occlusion masks, and Mixamo Skeleton. The dataset comprises 395 video sequences, totally 24,230 training frames, 720 validation frames, and 4,320 test frames. Furthermore, a comprehensive benchmark is constructed with various optical flow estimation methods to analyze the shortcomings and limitations across multiple datasets.
Problem
Research questions and friction points this paper is trying to address.
Lack of optical flow datasets for cel anime character motion
Need for high-quality dataset to support anime video generation
Absence of benchmarks for optical flow in anime domain
Innovation
Methods, ideas, or system contributions that make the work stand out.
3D model rendering for anime dataset
Rich annotations including optical flow
Comprehensive benchmark with various methods
🔎 Similar Papers
No similar papers found.
