Research map / Video / Video segmentation and tracking
Artistic Style Transfer for Videos and Spherical Images
Manuel Ruder, Alexey Dosovitskiy, Thomas Brox
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 |
|---|---|---|---|
| initialize with motion-compensated previous frame instead of unwarped previous frame (ablation) | optimization-based video style transfer · MPI Sintel alley_2 | temporal MSE: 0.010 -> 0.0016 | Improved |
| add explicit temporal penalty (ablation) | optimization-based video style transfer · MPI Sintel alley_2 | temporal MSE: 0.0016 -> 0.00061 | Improved |
| use DeepFlow instead of EpicFlow (ablation) | optimization-based video style transfer · MPI Sintel alley_2 | temporal MSE: 0.00073 -> 0.00061 | Improved |
| add dilated convolutions instead of strided downsampling (ablation) | network-based style transfer · MPI Sintel | temporal loss vs baseline: 100% -> 67% | Improved |
| train with two frames and a prior image (ablation) | network-based style transfer · MPI Sintel | temporal loss vs baseline: 100% -> 41% temporal; style 100% -> 167% | Mixed |
| add multi-frame training over two-frame training (ablation) | network-based style transfer · MPI Sintel | style loss vs baseline: 167% -> 85% | Improved |
| add mixed training (zero-prior samples) to multi-frame training (ablation) | network-based style transfer · MPI Sintel | style loss vs baseline: 85% -> 77% | Improved |
| replace independent per-frame network with prior-image network | network-based style transfer · MPI Sintel alley_2 | temporal MSE: 0.0062 -> 0.0016 | Improved |
| add prior image from neighboring cube face border | network-based style transfer · spherical images | cut-edge gradient E_grad_cut: 2.48 -> 1.40 | Improved |
| fine-tune video net on spherical-specific dataset (ablation) | network-based style transfer · spherical images | E_grad: 1.41 -> 1.21 | Improved |
| fill missing prior regions with random noise instead of zeros (ablation) | network-based style transfer · spherical images | E_grad: 1.21 -> 1.18 | Improved |
| add multi-pass algorithm (ablation) | optimization-based video style transfer · Miss Marple scene (qualitative) | image brightness and contrast over time | Improved |
Closest papers by meaning
- FlowStyler: Artistic Video Stylization via Transformation Fields Transports (ICCV 2025 · similarity 0.91)
- Multimodal Transfer: A Hierarchical Deep Convolutional Neural Network for Fast Artistic Style Transfer (CVPR 2017 · similarity 0.90)
- Artistic Style Transfer with Internal-external Learning and Contrastive Learning (NeurIPS 2021 · similarity 0.89)
- StyleMesh: Style Transfer for Indoor 3D Scene Reconstructions (CVPR 2022 · similarity 0.89)
- A Flexible Convolutional Solver for Fast Style Transfers (CVPR 2019 · similarity 0.89)
- ArtBank: Artistic Style Transfer with Pre-trained Diffusion Model and Implicit Style Prompt Bank (AAAI 2024 · similarity 0.87)
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.
