SPECSIA: Stylization Dataset for Novel-View Enhancement in Drawing-based 3D Animation
2026-07-01 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address the problem of animating a character from a single 2D drawing, which is hard because the animation needs to look consistent and natural from different angles. They found that current methods often get stuck fitting just one view and create errors when seen from other views. To fix this, the authors created a new dataset called SPECSIA-15K, which pairs problematic images with their corrected versions. They also developed DraViE, a simple tool that uses this data to clean up these errors in new views while keeping the character's style and movements believable. Tests showed their method improves animation quality across views with less effort than previous techniques.
2D animationcharacter appearancetemporal coherenceprojection artifacts3DBiCar datasetview consistencypaired stylization datasetsample-wise refinementDraViE modulenovel-view fidelity
Authors
Kyuwon Kim, Sunjae Yoon, Chang D. Yoo
Abstract
Generating animation from a single 2D drawing is challenging because the output must preserve character appearance while remaining plausible and temporally coherent under motion. Existing drawing-based 3D animation pipelines often use sample-wise 2D refinement to align animated renderings with the input image, but such optimization tends to overfit to the observed view and fails to correct projection-induced artifacts in novel views. To address this limitation, we introduce SPECSIA-15K, a paired stylization dataset containing 14,980 artifact-corrupted projection/refinement-target pairs from 1,498 3DBiCar characters. We further present DraViE (Drawing-based View Enhancement), a lightweight plug-and-play module trained with data-level priors to remove novel-view artifacts while preserving style and motion plausibility. Experiments show consistent gains in novel-view fidelity and temporal coherence with lower per-character adaptation cost than sample-wise fine-tuning.