Research map / 3D vision / Depth and stereo
S³Net: Semantic-Aware Self-supervised Depth Estimation with Monocular Videos and Synthetic Data
Bin Cheng, Inderjot Singh Saggu, Raunak Shah, Gaurav Bansal, Dinesh Bharadia
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 synthetic-to-real image translation (native) (ablation) | synthetic-supervised depth network · KITTI (50m cap) | Abs Rel: 0.278 -> 0.168 | Improved |
| add flow-guided photometric consistency to image translation (ablation) | native translation network · KITTI (50m cap) | Abs Rel: 0.168 -> 0.1539 | Improved |
| add semantic consistency loss to image translation (ablation) | native translation network · KITTI (50m cap) | Abs Rel: 0.168 -> 0.1544 | Improved |
| continue training S_seg instead of freezing it (ablation) | semantic consistency loss · KITTI (50m cap) | Abs Rel: 0.1555 -> 0.1544 | Improved |
| add semantic augmentation to depth network input only (ablation) | translated supervised model · KITTI (50m cap) | Abs Rel: 0.1544 -> 0.1532 | Improved |
| extend semantic augmentation to image translation network as well (ablation) | translated supervised model · KITTI (50m cap) | Abs Rel: 0.1532 -> 0.1455 | Improved |
| add real-world self-supervised photometric loss with joint training (ablation) | synthetic translated supervised model · KITTI (50m cap) | Abs Rel: 0.1455 -> 0.1292 | Improved |
| add bi-directional auto-masking (ablation) | self-supervised branch · KITTI (50m cap) | Abs Rel: 0.1292 -> 0.1198 | Improved |
| add semantic labels to input images in final model (ablation) | full S3Net · KITTI (50m cap) | Abs Rel: 0.1198 -> 0.1183 | Improved |
| add semantic augmentation (ablation) | S3Net · Make3D generalization | Abs Rel: 0.372 -> 0.322 | Improved |
| add semantic augmentation (ablation) | S3Net · Make3D generalization | Sq Rel: 5.699 -> 3.238 | Improved |
| alternating training instead of weighted-sum training of supervised and self-supervised losses (ablation) | S3Net depth network · KITTI | depth accuracy | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GeoNet — GeoNet: Geometric Neural Network for Joint Depth and Surface Normal Estimation (CVPR 2018) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.91)
- Self-Supervised Monocular Depth Estimation: Solving the Dynamic Object Problem by Semantic Guidance (ECCV 2020 · similarity 0.91)
- SC-DepthV3: Robust Self-Supervised Monocular Depth Estimation for Dynamic Scenes (TPAMI 2023 · similarity 0.91)
- 3D Packing for Self-Supervised Monocular Depth Estimation (CVPR 2020 · similarity 0.90)
- Self-Supervised Monocular Trained Depth Estimation Using Self-Attention and Discrete Disparity Volume (CVPR 2020 · similarity 0.90)
- HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation (AAAI 2021 · similarity 0.90)
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.
