Research map / Video / Video segmentation and tracking
MarioNette: Self-Supervised Sprite Learning
Dmitriy Smirnov, MICHAEL GHARBI, Matthew Fisher, Vitor Campagnolo Guizilini, Alexei A Efros, Justin Solomon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use MarioNette's shared sprite dictionary and layered sprite decomposition instead of SPACE | MarioNette · Synthetic sprite-based game | mean IoU | Improved |
| Use MarioNette's shared sprite dictionary and layered sprite decomposition instead of SPACE | MarioNette · Synthetic sprite-based game | binary foreground/background mean IoU | Improved |
| Use 16×16 patches instead of the full model's patch size (ablation) | MarioNette · Mario | reconstruction PSNR | Improved |
| Use 16×16 patches instead of the full model's patch size (ablation) | MarioNette · Mario | sprite decomposition consistency | Hurt |
| Use 64×64 patches instead of the full model's patch size (ablation) | MarioNette · Mario | reconstruction PSNR | Mixed |
| Reduce the dictionary to 25 elements (ablation) | MarioNette · Mario | reconstruction PSNR | Mixed |
| Remove LayerNorm (ablation) | MarioNette · Mario | reconstruction PSNR | Hurt |
| Replace Beta regularization on switches with a straight-through estimator for discrete switches (ablation) | MarioNette · Mario | reconstruction PSNR | No clear effect |
| Predict background shifts with a spatial transformer instead of discrete classification (ablation) | MarioNette · Mario | background recovery | Hurt |
Closest papers by meaning
- Fast Sprite Decomposition from Animated Graphics (ECCV 2024 · similarity 0.87)
- Deformable Sprites for Unsupervised Video Decomposition (CVPR 2022 · similarity 0.86)
- Unsupervised Layered Image Decomposition Into Object Prototypes (ICCV 2021 · similarity 0.85)
- AniDoc: Animation Creation Made Easier (CVPR 2025 · similarity 0.82)
- Infinite-Canvas: Higher-Resolution Video Outpainting with Extensive Content Generation (AAAI 2025 · similarity 0.82)
- EasyCraft: A Robust and Efficient Framework for Automatic Avatar Crafting (CVPR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
