Research map / 3D vision / Face and shape
MaRINeR: Enhancing Novel Views by Matching Rendered Images with Nearby References
Lukas Bösiger, Mihai Dusmanu, Marc Pollefeys, Zuria Bauer
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 |
|---|---|---|---|
| enhance renderings with MaRINeR versus unenhanced render | MaRINeR · LaMAR CAB | LPIPS: 0.380 -> 0.180 | Improved |
| add mesh down-sampling augmentation (10% triangles) in training (ablation) | MaRINeR · LaMAR CAB, 10% mesh | PSNR: 17.98 -> 19.10 | Improved |
| add random-reference-within-5s augmentation (ablation) | MaRINeR · LaMAR CAB, reference level 8 | PSNR: 16.79 -> 18.04 | Improved |
| add random-reference augmentation when reference is GT (ablation) | MaRINeR · LaMAR CAB, reference level 0 | PSNR: 23.51 -> 22.91 | Hurt |
| use iterative refinement with 2 iterations instead of 1 (ablation) | MaRINeR · LaMAR CAB | PSNR/SSIM | Improved |
| add SAM to decoder (ablation) | MaRINeR · LaMAR CAB | PSNR: 19.73 -> 19.88 | Improved |
| add DRAM to decoder (ablation) | MaRINeR · LaMAR CAB | PSNR: 19.71 -> 19.88 | Improved |
| use learned 64-channel encoder instead of pretrained VGG (ablation) | MaRINeR · LaMAR CAB | PSNR: 19.08 -> 19.88 | Improved |
| use larger-channel learned encoder (64 128 256) (ablation) | MaRINeR · LaMAR CAB | PSNR: 19.88 -> 19.66 | No clear effect |
| post-process nerfacto renderings with MaRINeR | nerfacto · NeRF postprocessing | LPIPS / PSNR: LPIPS 0.238 -> 0.167; PSNR 21.14 -> 20.45 | Mixed |
| enhance renderings before homography estimation | SuperPoint+SuperGlue homography · LaMAR | homography error: 4.86 -> 1.88 | Improved |
| raise adversarial loss weight to 0.005 (ablation) | MaRINeR · LaMAR | image quality | Hurt |
Closest papers by meaning
- SEAR: Simple and Efficient Adaptation of Visual Geometric Transformers for RGB+Thermal 3D Reconstruction (ECCV 2026 · similarity 0.86)
- Continuous 3D Perception Model with Persistent State (CVPR 2025 · similarity 0.84)
- MV-DUSt3R+: Single-Stage Scene Reconstruction from Sparse Views In 2 Seconds (CVPR 2025 · similarity 0.84)
- ViDAR: Video Diffusion-Aware 4D Reconstruction From Monocular Inputs (NeurIPS 2025 · similarity 0.84)
- WildCAT3D: Appearance-Aware Multi-View Diffusion in the Wild (NeurIPS 2025 · similarity 0.84)
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning (CVPR 2024 · similarity 0.83)
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.
