Research map / 3D vision / Depth and stereo
PanoFormer: Panorama Transformer for Indoor 360° Depth Estimation
Zhijie Shen, Chunyu Lin, Kang Liao, Lang Nie, Zishuo Zheng, Yao Zhao
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 |
|---|---|---|---|
| replace convolutions with Transformer (PST blocks) (ablation) | CNN baseline · Stanford2D3D (70 epochs), ablation | RMSE: 0.6704 -> 0.4349 | Improved |
| add tangent-patch division (STLM) vs standard attention (ablation) | Transformer b · Stanford2D3D (70 epochs), ablation | RMSE: 0.4349 -> 0.3739 | Improved |
| add tangent-patch division (STLM) (ablation) | Transformer b · Stanford2D3D (70 epochs), ablation | P-RMSE: 0.2068 -> 0.1825 | Improved |
| add token flow (ablation) | PanoFormer without token flow · Stanford2D3D (70 epochs), ablation | P-RMSE: 0.1825 -> 0.1793 | Improved |
| add token flow (ablation) | PanoFormer without token flow · Stanford2D3D (70 epochs), ablation | RMSE: 0.3739 -> 0.3366 | Improved |
| add token flow (ablation) | PanoFormer without token flow · Stanford2D3D (70 epochs), ablation | LRCE: 0.1916 -> 0.1784 | Improved |
| use PanoFormer instead of UniFuse | UniFuse · Matterport3D | RMSE: 0.4941 -> 0.3635 | Improved |
| use PanoFormer instead of SliceNet | SliceNet · Stanford2D3D | delta1: 0.9031 -> 0.9394 | Improved |
| use PanoFormer instead of DAMO | DAMO · 3D60 | RMSE: 0.1769 -> 0.1492 | Improved |
| apply unchanged PanoFormer to semantic segmentation | HoHoNet · Stanford2D3D semantic segmentation | mIoU: 43.3 -> 48.9 | Improved |
Later papers that compared with this method
3 later accepted papers report PanoFormer as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- CasaMaestro: Multi-View Panoramas for House-Scale 3D Reconstruction (ECCV 2026 · better on 3 of 3 shared results)
- EGformer: Equirectangular Geometry-biased Transformer for 360 Depth Estimation (ICCV 2023 · better on 7 of 7 shared results)
- HRDFuse: Monocular 360deg Depth Estimation by Collaboratively Learning Holistic-With-Regional Depth Distributions (CVPR 2023 · better on 0 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SliceNet — SliceNet: Deep Dense Depth Estimation From a Single Indoor Panorama Using a Slice-Based Representation (CVPR 2021) · better on 2 of 2 shared results
- Bifuse — BiFuse++: Self-Supervised and Efficient Bi-Projection Fusion for $360^{\circ }$ Depth Estimation (TPAMI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Geometric Exploitation for Indoor Panoramic Semantic Segmentation (NeurIPS 2024 · similarity 0.89)
- Unified Panoramic Geometry Estimation via Multi-View Foundation Models (NeurIPS 2026 · similarity 0.86)
- PanoContext-Former: Panoramic Total Scene Understanding with a Transformer (CVPR 2024 · similarity 0.86)
- PanDA: Towards Panoramic Depth Anything with Unlabeled Panoramas and Mobius Spatial Augmentation (CVPR 2025 · similarity 0.86)
- Splatter-360: Generalizable 360 Gaussian Splatting for Wide-baseline Panoramic Images (CVPR 2025 · similarity 0.85)
- DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images (NeurIPS 2026 · 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.
