SPECSIA: Stylization Dataset for Novel-View Enhancement in Drawing-based 3D Animation

2026-07-01

AI Summary

Existing single-image-based 3D animation methods often suffer from projection artifacts under novel views and struggle to simultaneously preserve stylistic consistency and temporal coherence. To address these limitations, this work introduces SPECSIA-15K, the first paired stylized dataset specifically designed for novel-view enhancement, comprising 14,980 image pairs. Furthermore, we propose DraViE, a lightweight, plug-and-play module that leverages data-driven priors to effectively suppress artifacts and maintain both visual style and motion plausibility without requiring per-sample fine-tuning. Experimental results demonstrate that our approach significantly outperforms baseline methods in terms of novel-view fidelity and temporal consistency, while substantially reducing the cost of character adaptation.

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.

Problem

• novel-view enhancement
• drawing-based 3D animation
• projection artifacts
• temporal coherence
• style preservation

Innovation

• novel-view enhancement
• drawing-based 3D animation
• artifact correction
• stylization dataset
• temporal coherence