Research map / 3D vision / Depth and stereo
Mono-ViFI: A Unified Learning Framework for Self-supervised Single- and Multi-frame Monocular Depth Estimation
Jinfeng Liu, Lingtong Kong, Bo Li, Zerong Wang, Hong Gu, Jinwei Chen
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 temporal augmentation (VFI-synthesized virtual views) (ablation) | Monodepth2 single-frame · Cityscapes | Abs Rel: 0.129 -> 0.119 | Improved |
| add scale-aware depth consistency on top of spatial augmentation (ablation) | Monodepth2 single-frame · Cityscapes | Abs Rel: 0.122 -> 0.116 | Improved |
| add standard view depth consistency (ablation) | Monodepth2 single-frame · Cityscapes | Abs Rel: 0.129 -> 0.117 | Improved |
| add motion-aware feature alignment instead of direct feature fusion (ablation) | multi-frame fusion · Cityscapes | Abs Rel: 0.120 -> 0.114 | Improved |
| add occlusion-alleviated feature fusion (ablation) | multi-frame fusion · Cityscapes | Abs Rel: 0.114 -> 0.111 | Improved |
| combine all components (SA, TA, SVDC, SADC) with fusion (ablation) | multi-frame model · Cityscapes | Abs Rel: 0.111 -> 0.107 | Improved |
| swap VFI network among IFRNet, RIFE, AMT (ablation) | Mono-ViFI multi-frame · KITTI | Abs Rel: 0.099 / 0.100 / 0.099 | No clear effect |
| remove Fourier positional encoding of optical flow (ablation) | multi-frame fusion · KITTI and Cityscapes | depth metrics | Hurt |
| share both encoder and decoder between single- and multi-frame models (ablation) | Mono-ViFI · KITTI and Cityscapes | depth metrics | Hurt |
| fully separate single- and multi-frame models instead of sharing the encoder (ablation) | Mono-ViFI · KITTI and Cityscapes | depth metrics | No clear effect |
| use large IFRNet in training, small in evaluation (ablation) | Mono-ViFI multi-frame · KITTI | depth metrics | No clear effect |
| apply Mono-ViFI to Monodepth2 ResNet18 single-frame inference | Monodepth2 · KITTI raw GT | Abs Rel: 0.115 -> 0.105 | Improved |
Closest papers by meaning
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.90)
- Multi-Frame Self-Supervised Depth With Transformers (CVPR 2022 · similarity 0.89)
- Self-Supervised Monocular Trained Depth Estimation Using Self-Attention and Discrete Disparity Volume (CVPR 2020 · similarity 0.89)
- Self-Supervised Multi-Frame Monocular Scene Flow (CVPR 2021 · similarity 0.89)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.89)
- Mining Supervision for Dynamic Regions in Self-Supervised Monocular Depth Estimation (CVPR 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.
