Research map / 3D vision / Depth and stereo
$SE(3)$ Equivariant Ray Embeddings for Implicit Multi-View Depth Estimation
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 |
|---|---|---|---|
| add SE(3) equivariance to the Perceiver IO architecture | DeFiNe on ScanNet · ScanNet | Abs.Rel.: 0.093 -> 0.086 | Helped |
| add SE(3) equivariance to the Perceiver IO architecture | DeFiNe on SUN3D · SUN3D | Abs.Rel.: 0.099 -> 0.090 | Helped |
| add SE(3) equivariance to the Perceiver IO architecture | IIB on RGBD-SLAM · RGBD-SLAM | RMSE: 0.550 -> 0.433 | Helped |
| add SE(3) equivariance to the Perceiver IO architecture | NAS on Scenes11 · Scenes11 | Abs.Rel.: 0.038 -> 0.069 | Hurt |
| add camera information (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.229 -> 0.086 | Helped |
| use spherical harmonics instead of Fourier encodings (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.131 -> 0.086 | Helped |
| use equivariant attention instead of conventional attention (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.127 -> 0.086 | Helped |
| increase maximum spherical harmonics order from 1 to 2 (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.134 -> 0.125 | Helped |
| increase maximum spherical harmonics order from 2 to 4 (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.125 -> 0.116 | Helped |
| use a learned canonical frame and invariant decoder instead of an equivariant decoder (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.128 -> 0.086 | Helped |
| add virtual camera augmentation (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.104 -> 0.086 | Helped |
| add canonical camera jittering (ablation) | EPIO on ScanNet · ScanNet | Abs.Rel.: 0.086 -> 0.086 | No clear effect |
Closest papers by meaning
- $SE(3)$ Equivariant Convolution and Transformer in Ray Space (NeurIPS 2023 · similarity 0.93)
- Multiview Equivariance Improves 3D Correspondence Understanding with Minimal Feature Finetuning (ICLR 2025 · similarity 0.89)
- Equivariant Point Network for 3D Point Cloud Analysis (CVPR 2021 · similarity 0.88)
- $\mathrm{SE}(3)$-Equivariant Attention Networks for Shape Reconstruction in Function Space (ICLR 2023 · similarity 0.88)
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction (NeurIPS 2024 · similarity 0.88)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · 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.