Research map / 3D vision / Depth and stereo
MGNet: Monocular Geometric Scene Understanding for Autonomous Driving
Markus Schön, Michael Buchholz, Klaus Dietmayer
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 |
|---|---|---|---|
| train depth and panoptic segmentation jointly with fixed weights (vs two single-task models) (ablation) | MGNet · Cityscapes val | PQ: 54.1 -> 53.3 | Hurt |
| train depth and panoptic segmentation jointly with fixed weights (vs two single-task models) (ablation) | MGNet · Cityscapes val | RMSE: 9.3 -> 8.8 | Improved |
| add homoscedastic uncertainty weighting with fixed and learnable terms (ablation) | MGNet multi-task baseline · Cityscapes val | PQ: 53.3 -> 54.2 | Improved |
| add homoscedastic uncertainty weighting with fixed and learnable terms (ablation) | MGNet multi-task baseline · Cityscapes val | RMSE: 8.8 -> 8.6 | Improved |
| add video sequence training with pseudo labels (ablation) | MGNet + uncertainty · Cityscapes val | PQ: 54.2 -> 55.7 | Improved |
| add video sequence training with pseudo labels (ablation) | MGNet + uncertainty · Cityscapes val | RMSE: 8.6 -> 8.3 | Improved |
| use ResNet18 backbone instead of MobileNetV3 (ablation) | MGNet · Cityscapes val | PQ: 48.6 -> 53.3 | Improved |
| use ResNet18 backbone instead of EfficientNetLite0 (ablation) | MGNet · Cityscapes val | RMSE: 8.6 -> 8.8 | Hurt |
| use ResNet18 backbone instead of EfficientNetLite0 (ablation) | MGNet · Cityscapes val | FPS: 27.7 -> 30.0 | Improved |
| replace bilateral spatial/context path encoding with a skip connection for low-level features (ablation) | MGNet decoder · Cityscapes | latency | Improved |
Closest papers by meaning
- 3D Packing for Self-Supervised Monocular Depth Estimation (CVPR 2020 · similarity 0.90)
- Towards Scene Understanding: Unsupervised Monocular Depth Estimation With Semantic-Aware Representation (CVPR 2019 · similarity 0.89)
- SQLdepth: Generalizable Self-Supervised Fine-Structured Monocular Depth Estimation (AAAI 2024 · similarity 0.89)
- S³Net: Semantic-Aware Self-supervised Depth Estimation with Monocular Videos and Synthetic Data (ECCV 2020 · similarity 0.88)
- The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth (CVPR 2021 · similarity 0.88)
- Geometry-Guided Domain Generalization for Monocular 3D Object Detection (AAAI 2024 · similarity 0.87)
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.
