Research map / 3D vision / Depth and stereo
Progressive Fusion for Unsupervised Binocular Depth Estimation Using Cycled Networks
Andrea Pilzer, Stéphane Lathuilière, Dan Xu, Mihai Puscas, Elisa Ricci, Nicu Sebe
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 the cycle (second half-cycle) to Half-Cycle + L_gan (ablation) | Half-Cycle Stereo · KITTI Eigen split (80m) | Abs Rel: 0.148 -> 0.137 | Improved |
| add feature warping (d+xi) on top of disparity-only warping (ablation) | Half-Cycle Stereo · KITTI Eigen split (80m) | Abs Rel: 0.159 -> 0.153 | Improved |
| add feature warping (d+xi) on top of disparity-only warping (ablation) | Cycle Stereo · KITTI Eigen split (80m) | Abs Rel: 0.146 -> 0.141 | Improved |
| add adversarial loss L_gan (ablation) | Cycle Stereo · KITTI Eigen split (80m) | Abs Rel: 0.141 -> 0.137 | Improved |
| add SSIM loss (ablation) | Cycle Stereo + L_gan · KITTI Eigen split (80m) | Abs Rel: 0.137 -> 0.102 | Improved |
| add cycle to feature-warped half-cycle (ablation) | Half-Cycle Stereo · Cityscapes | Abs Rel: 0.436 -> 0.420 | Improved |
| add SSIM loss (ablation) | Cycle Stereo + L_gan · Cityscapes | Abs Rel: 0.418 -> 0.404 | Improved |
| add SSIM loss (ablation) | Cycle + L_gan · ApolloScape | Abs Rel: 0.423 -> 0.387 | Improved |
| add adversarial loss L_gan (ablation) | Half-Cycle · ApolloScape | Abs Rel: 0.469 -> 0.446 | Improved |
| add feature and disparity alignment (versus plain U-Net concatenation) (ablation) | Half-Cycle d+xi · KITTI Eigen split (80m) | Abs Rel: 0.159 -> 0.154 | Improved |
| use separate discriminators D_r and D_l instead of one shared (ablation) | Half-Cycle + L_gan · KITTI Eigen split (80m) | Abs Rel: 0.155 -> 0.148 | Improved |
| feed both resynthesized images (I_r hat, I_l hat) to the second half-cycle (ablation) | Cycle + L_gan · KITTI Eigen split (80m) | Abs Rel: 0.146 -> 0.137 | Improved |
Closest papers by meaning
- Bilateral Cyclic Constraint and Adaptive Regularization for Unsupervised Monocular Depth Prediction (CVPR 2019 · similarity 0.90)
- Refine and Distill: Exploiting Cycle-Inconsistency and Knowledge Distillation for Unsupervised Monocular Depth Estimation (CVPR 2019 · similarity 0.89)
- Du²Net: Learning Depth Estimation from Dual-Cameras and Dual-Pixels (ECCV 2020 · similarity 0.89)
- Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue (ECCV 2016 · similarity 0.89)
- Unsupervised Monocular Depth Estimation With Left-Right Consistency (CVPR 2017 · similarity 0.89)
- Self-Supervised Monocular Depth Estimation by Direction-aware Cumulative Convolution Network (ICCV 2023 · similarity 0.89)
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.
