Research map / 3D vision / Depth and stereo
BiFuse++: Self-Supervised and Efficient Bi-Projection Fusion for $360^{\circ }$ Depth Estimation
Fu-En Wang, Yu-Hsuan Yeh, Yi-Hsuan Tsai, Wei-Chen Chiu, Min Sun
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 |
|---|---|---|---|
| use BiFuse++ instead of BiFuse | depth fusion · Matterport3D | MAE: 0.3470 -> 0.2842 | Improved |
| use BiFuse++ instead of HoHoNet | depth estimation · Matterport3D | MAE: 0.2862 -> 0.2842 | No clear effect |
| use BiFuse++ instead of HoHoNet | depth estimation · Stanford2D3D | MAE: 0.2027 -> 0.2173 | Hurt |
| use BiFuse++ instead of UniFuse | depth estimation · PanoSUNCG (supervised) | MAE: 0.0776 -> 0.0688 | Improved |
| use BiFuse++ instead of HoHoNet under rotation noise | depth estimation · Matterport3D (rotation noise) | MAE: 0.3085 -> 0.3054 | Improved |
| use BiFuse++ instead of HoHoNet under rotation noise | depth estimation · Stanford2D3D (rotation noise) | MAE: 0.2210 -> 0.2193 | Improved |
| use new fusion module instead of BiFuse module | fusion module · fusion module parameters | parameters: 4.7 M -> 2.1 M | Improved |
| use BiFuse++ instead of BiFuse | full network · GFLOPs / runtime memory | GFLOPs: 682.86 -> 87.42 GFLOPs; 3346 -> 1762 Mb | Improved |
| add contrast-aware photometric loss (CAPL) instead of SPL (ablation) | BiFuse++ · PanoSUNCG (self-supervised) | RMSE: 0.4695 -> 0.4321 | Improved |
| use SPL instead of SSIM photometric loss (ablation) | BiFuse++ · PanoSUNCG (self-supervised) | RMSE: 0.5889 -> 0.4695 | Improved |
| use BiFuse++ w/ CAPL instead of 360-SelfNet | self-supervised depth · PanoSUNCG (self-supervised) | RMSE: 0.5121 -> 0.4321 | Improved |
| use BiFuse++ w/ CAPL instead of UniFuse | self-supervised depth · PanoSUNCG (self-supervised) | MAE: 0.2452 -> 0.1815 | Improved |
Later papers that compared with this method
5 later accepted papers report Bifusev2 as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- PanoVGGT: Feed-Forward 3D Reconstruction from Panoramic Imagery (CVPR 2026 · better on 2 of 2 shared results)
- UniK3D: Universal Camera Monocular 3D Estimation (CVPR 2025 · better on 1 of 1 shared result)
- Depth Anywhere: Enhancing 360 Monocular Depth Estimation via Perspective Distillation and Unlabeled Data Augmentation (NeurIPS 2024 · better on 2 of 2 shared results)
- EGformer: Equirectangular Geometry-biased Transformer for 360 Depth Estimation (ICCV 2023 · better on 1 of 1 shared result)
- PanoFormer: Panorama Transformer for Indoor 360° Depth Estimation (ECCV 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
- BiFuse: Monocular 360 Depth Estimation via Bi-Projection Fusion (CVPR 2020 · similarity 0.91)
- Improving 360 Monocular Depth Estimation via Non-local Dense Prediction Transformer and Joint Supervised and Self-Supervised Learning (AAAI 2022 · similarity 0.89)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.87)
- Self-Supervised Monocular Trained Depth Estimation Using Self-Attention and Discrete Disparity Volume (CVPR 2020 · similarity 0.87)
- CroMo: Cross-Modal Learning for Monocular Depth Estimation (CVPR 2022 · similarity 0.86)
- Elite360D: Towards Efficient 360 Depth Estimation via Semantic- and Distance-Aware Bi-Projection Fusion (CVPR 2024 · 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.
