Research map / 3D vision / NeRF and radiance fields
Cross-Ray Neural Radiance Fields for Novel-View Synthesis from Unconstrained Image Collections
Yifan Yang, Shuhai Zhang, Zixiong Huang, Yubing Zhang, Mingkui Tan
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 cross-ray appearance modeling module (ablation) | CR-NeRF · Brandenburg Gate (R/4) | PSNR: 20.46 -> 26.86 | Improved |
| add transient handling module (ablation) | CR-NeRF · Brandenburg Gate (R/4) | PSNR: 26.38 -> 26.86 | Improved |
| use cross-ray features instead of ray-point-level features (ablation) | CR-NeRF · Brandenburg Gate (R/4) | PSNR: 26.18 -> 26.86 | Improved |
| use pre-trained DeepLabV3 to mask common transient objects (ablation) | CR-NeRF-A · Brandenburg Gate (R/4) | PSNR: 26.38 -> 24.89 | Hurt |
| use pre-trained Mask R-CNN to mask common transient objects (ablation) | CR-NeRF-A · Brandenburg Gate (R/4) | PSNR: 26.38 -> 25.46 | Hurt |
| use learned visibility segmentation network instead of NeRF-W transient network (ablation) | CR-NeRF · Brandenburg Gate (R/4) | PSNR: 24.72 -> 26.86 | Improved |
| increase number of sampled rays beyond 784 (ablation) | CR-NeRF · Brandenburg Gate (R/4) | PSNR: 27.12 -> 26.69 (784 -> 1600 rays) | Hurt |
| use cross-ray manner instead of single-ray (vs Ha-NeRF) across ray counts (ablation) | CR-NeRF · Brandenburg Gate PSNR vs number of rays | PSNR: 24.51 -> 26.86 at 1024 rays | Improved |
| generate cross-ray features once and fuse with each appearance embedding | CR-NeRF · Brandenburg Gate inference over multiple appearances | inference time (s): 24.09 -> 2.12 | Improved |
| CR-NeRF versus NeRF-W on Sacre Coeur SSIM | CR-NeRF · Sacre Coeur (R/4) | SSIM: 0.8369 vs 0.8422 (NeRF-W) | Mixed |
Closest papers by meaning
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.92)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.92)
- Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields (CVPR 2022 · similarity 0.92)
- Neural Transmitted Radiance Fields (NeurIPS 2022 · similarity 0.91)
- NeRF-VPT: Learning Novel View Representations with Neural Radiance Fields via View Prompt Tuning (AAAI 2024 · similarity 0.91)
- Few-Shot Neural Radiance Fields under Unconstrained Illumination (AAAI 2024 · similarity 0.91)
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.
