Research map / 3D vision / Depth and stereo
GasMono: Geometry-Aided Self-Supervised Monocular Depth Estimation for Indoor Scenes
Chaoqiang Zhao, Matteo Poggi, Fabio Tosi, Lei Zhou, Qiyu Sun, Yang Tang, Stefano Mattoccia
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 ResNet18 encoder with MPViT transformer (ablation) | Monodepth2 · NYUv2 | Abs Rel: 0.167 -> 0.145 | Improved |
| use raw COLMAP poses in place of PoseNet (ablation) | Monodepth2 · NYUv2 | Abs Rel: 0.167 -> 0.298 | Hurt |
| keep COLMAP rotation but use network-estimated translation (ablation) | Monodepth2 · NYUv2 | Abs Rel: 0.167 -> 0.160 | Improved |
| learn only a scale factor for COLMAP translation (ablation) | COLMAP poses · NYUv2 | scale std: 0.597 -> 0.237 | Improved |
| learn only residual shift on COLMAP translation (ablation) | COLMAP poses · NYUv2 | Abs Rel: 0.298 -> 0.255 | No clear effect |
| learn both scale and residual shift of translation (ablation) | COLMAP poses · NYUv2 | Abs Rel: 0.298 -> 0.159 | Improved |
| add translation optimization (AlignNet) (ablation) | MPViT + COLMAP poses · NYUv2 | Abs Rel: 0.311 -> 0.124 | Improved |
| add rotation optimization (residual PoseNet) (ablation) | MPViT + COLMAP + Optim_t · NYUv2 | Abs Rel: 0.124 -> 0.114 | Improved |
| add iterative self-distillation (ISD) to MPViT model (ablation) | MPViT · NYUv2 | Abs Rel: 0.145 -> 0.137 | Improved |
| add ISD to full pose-optimized model (ablation) | GasMono · NYUv2 | Abs Rel: 0.114 -> 0.113 | Improved |
| add ISD instead of naive SD (ablation) | Monodepth2 · KITTI | Abs Rel: 0.112 (SD) -> 0.111 (ISD) | Improved |
| add naive SD to MonoViT (ablation) | MonoViT · KITTI | Abs Rel: 0.099 -> 0.099 | No clear effect |
Closest papers by meaning
- StructDepth: Leveraging the Structural Regularities for Self-Supervised Indoor Depth Estimation (ICCV 2021 · similarity 0.90)
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.89)
- Fine-Grained Semantics-Aware Representation Enhancement for Self-Supervised Monocular Depth Estimation (ICCV 2021 · similarity 0.89)
- Toward Practical Monocular Indoor Depth Estimation (CVPR 2022 · similarity 0.88)
- S³Net: Semantic-Aware Self-supervised Depth Estimation with Monocular Videos and Synthetic Data (ECCV 2020 · similarity 0.88)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 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.
