Research map / 3D vision / Depth and stereo
Stereoscopic Neural Style Transfer
Dongdong Chen, Lu Yuan, Jing Liao, Nenghai Yu, Gang Hua
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 |
|---|---|---|---|
| add symmetric bidirectional disparity consistency loss | stereoscopic style transfer · stereoscopic images | disparity consistency | Improved |
| jointly process the left and right views symmetrically rather than stylizing sequentially | optimization-based stereoscopic style transfer · stereoscopic images | texture continuity near occlusion boundaries | Improved |
| use bidirectional rather than single-directional disparity and occlusion prediction (ablation) | DispOccNet · MPI Sintel clean | non-occluded EPE: 4.66 -> 4.16 | Improved |
| use bidirectional rather than single-directional disparity and occlusion prediction (ablation) | DispOccNet · FlyingThings3D test | non-occluded EPE: 1.69 -> 1.68 | No clear effect |
| remove disparity deviation penalty in occluded regions (ablation) | DispOccNet · MPI Sintel clean | non-occluded EPE: 5.43 -> 4.16 | Improved |
| remove disparity deviation penalty in occluded regions (ablation) | DispOccNet · FlyingThings3D test | non-occluded EPE: 1.99 -> 1.68 | Improved |
| predict occlusion masks directly rather than derive them by post bidirectional consistency check | DispOccNet · FlyingThings3D test | occlusion-mask F-score: 0.805 -> 0.887 | Improved |
| add DispOccNet and the disparity consistency training to the style network | stereoscopic style transfer · FlyingThings3D test, Candy | disparity loss: 0.0624 -> 0.0474 | Improved |
| add DispOccNet and the disparity consistency training to the style network | stereoscopic style transfer · FlyingThings3D test, La_muse | disparity loss: 0.0403 -> 0.0301 | Improved |
| add DispOccNet and the disparity consistency training to the style network | stereoscopic style transfer · FlyingThings3D test, Mosaic | disparity loss: 0.0668 -> 0.0559 | Improved |
| add DispOccNet and the disparity consistency training to the style network | stereoscopic style transfer · FlyingThings3D test, Udnie | disparity loss: 0.0379 -> 0.0285 | Improved |
| replace symmetric bidirectional feature composition with asymmetric single-directional composition (ablation) | stereoscopic style transfer · FlyingThings3D test, Candy | disparity loss: 0.0474 -> 0.0481 | Hurt |
Closest papers by meaning
- Neural Stereoscopic Image Style Transfer (ECCV 2018 · similarity 0.94)
- DiffStyle3D: Consistent 3D Gaussian Stylization via Attention Optimization (ICML 2026 · similarity 0.87)
- Artistic Style Transfer for Videos and Spherical Images (IJCV 2018 · similarity 0.87)
- FPRF: Feed-Forward Photorealistic Style Transfer of Large-Scale 3D Neural Radiance Fields (AAAI 2024 · similarity 0.86)
- Styl3R: Instant 3D Stylized Reconstruction for Arbitrary Scenes and Styles (NeurIPS 2025 · similarity 0.85)
- 3DToonify: Creating Your High-Fidelity 3D Stylized Avatar Easily from 2D Portrait Images (CVPR 2024 · 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.
