Research map / Label-efficient and robust learning / Self-supervised and contrastive
Feature-metric Loss for Self-supervised Learning of Depth and Egomotion
Chang Shu, Kun Yu, Zhixiang Duan, Kuiyuan Yang
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 smoothness losses with feature-metric loss (ablation) | Lph · KITTI Eigen (MS, no online refinement) | Sq Rel: 0.735 -> 0.697 | Improved |
| replace smoothness losses with feature-metric loss, with online refinement (ablation) | Lph · KITTI Eigen (MS, online refinement) | Abs Rel: 0.088 -> 0.079 | Improved |
| add online refinement to feature-metric loss (ablation) | Lph + Lfm · KITTI Eigen (MS) | Abs Rel: 0.099 -> 0.079 | Improved |
| add online refinement to smoothness losses (ablation) | Lph + L1ds + L2ds · KITTI Eigen (MS) | Abs Rel: 0.103 -> 0.088 | Improved |
| add second-order smoothness to first-order smoothness (ablation) | Lph + L1ds · KITTI Eigen (MS, no online refinement) | Abs Rel: 0.105 -> 0.103 | Improved |
| add discriminative loss to feature learning (ablation) | Lrec · KITTI Eigen (FeatureNet loss ablation) | Abs Rel: 0.105 -> 0.103 | Improved |
| add convergent loss to feature learning (ablation) | Lrec · KITTI Eigen (FeatureNet loss ablation) | Abs Rel: 0.105 -> 0.100 | Improved |
| add both discriminative and convergent losses (ablation) | Lrec · KITTI Eigen (FeatureNet loss ablation) | Abs Rel: 0.105 -> 0.099 | Improved |
| add both discriminative and convergent losses (ablation) | Lrec · KITTI odometry Seq. 09 | terr: 4.30 -> 3.07 | Improved |
| add both discriminative and convergent losses (ablation) | Lrec · KITTI odometry Seq. 10 | rerr: 4.06 -> 1.78 | Improved |
| add discriminative loss on top of convergent loss (ablation) | Lrec + Lcvt · KITTI Eigen (FeatureNet loss ablation) | Abs Rel: 0.100 -> 0.099 | Improved |
Closest papers by meaning
- Unsupervised Learning of Monocular Depth Estimation and Visual Odometry With Deep Feature Reconstruction (CVPR 2018 · similarity 0.88)
- Digging Into Self-Supervised Monocular Depth Estimation (ICCV 2019 · similarity 0.87)
- Sequential Adversarial Learning for Self-Supervised Deep Visual Odometry (ICCV 2019 · similarity 0.87)
- Self-Supervised Monocular Depth Hints (ICCV 2019 · similarity 0.85)
- Deep Virtual Stereo Odometry: Leveraging Deep Depth Prediction for Monocular Direct Sparse Odometry (ECCV 2018 · similarity 0.85)
- Towards Better Generalization: Joint Depth-Pose Learning Without PoseNet (CVPR 2020 · 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.
