Research map / Video / Video generation
Consistent Video Style Transfer via Compound Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| add global feature sharing across the sequence (ablation) | Baseline style transfer network · MPI Sintel, interval i=1 | temporal loss: 0.059 -> 0.050 | Helped |
| add compound temporal regularization Lt (ablation) | Baseline style transfer network · MPI Sintel, interval i=16 | temporal loss: 0.063 -> 0.050 | Helped |
| add global feature sharing on top of Lt (ablation) | Baseline + Lt · MPI Sintel, interval i=1 | temporal loss: 0.041 -> 0.036 | Helped |
| add Blind post-processing (ablation) | Baseline · MPI Sintel, interval i=16 | temporal loss: 0.063 -> 0.063 | No clear effect |
| use compound regularization Lcomp instead of noise-only or motion-only regularization (ablation) | baseline style transfer network · temporal vs style loss trade-off | style loss at given temporal loss | Helped |
| use Lcomp instead of Lmotion for content loss (ablation) | baseline style transfer network · temporal vs content loss trade-off | content loss | Mixed |
| train with compound regularization instead of directly on real videos (ablation) | baseline style transfer network · temporal vs style loss trade-off | style/temporal loss | Helped |
| make Filter Predictor dynamic to both content and style instead of style only (ablation) | dynamic inter-channel filter · image style transfer (qualitative) | stylization quality | Helped |
| increase temporal regularization weight (ablation) | style transfer network · temporal vs style loss trade-off | style loss | Hurt |
Closest papers by meaning
- Real-Time Neural Style Transfer for Videos (CVPR 2017 · similarity 0.91)
- Learning Linear Transformations for Fast Image and Video Style Transfer (CVPR 2019 · similarity 0.89)
- Two-Stage Peer-Regularized Feature Recombination for Arbitrary Image Style Transfer (CVPR 2020 · similarity 0.87)
- Dynamic Instance Normalization for Arbitrary Style Transfer (AAAI 2020 · similarity 0.87)
- Semantix: An Energy-guided Sampler for Semantic Style Transfer (ICLR 2025 · similarity 0.87)
- StyleMaster: Stylize Your Video with Artistic Generation and Translation (CVPR 2025 · similarity 0.85)
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.