Research map / 3D vision / NeRF and radiance fields
Behind the Scenes: Density Fields for Single View Reconstruction
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 |
|---|---|---|---|
| Train with stereo and fisheye views rather than removing them (ablation) | Ours · KITTI-360 | IE_acc: 0.70 -> 0.77 | Helped |
| Train with stereo and fisheye views rather than removing them (ablation) | Ours · KITTI-360 | IE_rec: 0.06 -> 0.43 | Helped |
| Use a powerful encoder-decoder and a small MLP instead of an encoder and big MLP (ablation) | Density-field model · KITTI-360 | IE_rec: 0.41 -> 0.43 | Mixed |
| Predict density only and sample color instead of predicting density and color (ablation) | Encoder-decoder, small-MLP model · KITTI-360 | IE_rec: 0.15 -> 0.43 | Helped |
| Replace the encoder with an encoder-decoder (ablation) | Small-MLP, density-only model · KITTI, Eigen split | RMSE: 4.590 -> 4.407 | Helped |
| Keep invalid rays instead of discarding them (ablation) | Ours · KITTI, Eigen split | RMSE: 4.407 -> 4.493 | Hurt |
| Keep invalid rays instead of discarding them (ablation) | Ours · KITTI-360 | IE_rec: 0.43 -> 0.41 | Hurt |
| Add fine importance sampling (ablation) | Ours · Reconstruction experiments | reconstruction quality | No clear effect |
| Use the full view setup instead of training without stereo and fisheye views (ablation) | Ours · KITTI-360 | IE_acc: 0.70 -> 0.77 | Helped |
| Use the full view setup instead of training without fisheye views (ablation) | Ours · KITTI-360 | IE_acc: 0.71 -> 0.77 | Helped |
Closest papers by meaning
- S$^3$-NeRF: Neural Reflectance Field from Shading and Shadow under a Single Viewpoint (NeurIPS 2022 · similarity 0.88)
- Towards Multimodal Depth Estimation From Light Fields (CVPR 2022 · similarity 0.88)
- Stereo Radiance Fields (SRF): Learning View Synthesis for Sparse Views of Novel Scenes (CVPR 2021 · similarity 0.88)
- Neural Point Light Fields (CVPR 2022 · similarity 0.88)
- Boosting Self-Supervision for Single-View Scene Completion via Knowledge Distillation (CVPR 2024 · similarity 0.88)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 2024 · similarity 0.88)
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.