Research map / 3D vision / Depth and stereo
Recurrent Neural Network for (Un-)Supervised Learning of Monocular Video Visual Odometry and Depth
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 multi-view recurrent depth estimation instead of single-view depth estimation | Ours-sup · KITTI Eigen Split | RMSE: 1.949 -> 1.698 | Helped |
| train with unsupervised multi-view constraints | Ours-unsup · KITTI Eigen Split | RMSE: 2.320 | Helped |
| use supervised multi-view depth and pose estimation | Ours · KITTI Odometry Seq 09 and Seq 10 | translation error: 9.30; 7.21 | Mixed |
| place ConvLSTM units only in the encoder instead of throughout the network or only in the decoder (ablation) | Depth estimation network · KITTI Eigen Split | RMSE: 1.764 -> 1.698; 1.808 -> 1.698 | Helped |
| add forward-backward flow consistency to consecutive image reprojection (ablation) | Supervised Ours-d · KITTI Eigen Split | RMSE: 1.785 -> 1.759 | Helped |
| replace consecutive reprojection with multi-view reprojection while retaining flow consistency (ablation) | Supervised Ours-dc · KITTI Eigen Split | RMSE: 1.759 -> 1.698 | Helped |
| replace consecutive reprojection with multi-view reprojection while retaining flow consistency (ablation) | Unsupervised Ours-dc · KITTI Eigen Split | RMSE: 2.689 -> 2.361 | Helped |
| increase temporal context from a single frame to 3 frames (ablation) | Ours-sup · KITTI Eigen Split | RMSE: 1.949 -> 1.707 | Helped |
| increase temporal context beyond 10 frames (ablation) | Ours-sup · KITTI Eigen Split | RMSE: 1.698 -> 1.711 | Hurt |
| run the recurrent estimator on the whole sequence rather than a 10-frame window (ablation) | Ours-sup · KITTI Eigen Split | RMSE: 1.698 -> 1.748 | Hurt |
Closest papers by meaning
- Unsupervised Learning of Monocular Depth Estimation and Visual Odometry With Deep Feature Reconstruction (CVPR 2018 · similarity 0.91)
- Learning Depth From Monocular Videos Using Direct Methods (CVPR 2018 · similarity 0.89)
- Depth From Camera Motion and Object Detection (CVPR 2021 · similarity 0.88)
- D3VO: Deep Depth, Deep Pose and Deep Uncertainty for Monocular Visual Odometry (CVPR 2020 · similarity 0.88)
- RIAV-MVS: Recurrent-Indexing an Asymmetric Volume for Multi-View Stereo (CVPR 2023 · similarity 0.88)
- CNN-SLAM: Real-Time Dense Monocular SLAM With Learned Depth Prediction (CVPR 2017 · 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.