Research map / 3D vision / 3D generation
Veritatem Dies Aperit - Temporally Consistent Depth Prediction Enabled by a Multi-Task Geometric and Semantic Scene Understanding Approach
Amir Atapour-Abarghouei, Toby P. Breckon
Check 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 |
|---|---|---|---|
| jointly train depth prediction and segmentation in one network instead of separate networks (ablation) | multi-task scene understanding model · Random synthetic sequences | Abs. Rel.: 0.245 -> 0.208 | Improved |
| jointly train depth prediction and segmentation in one network instead of separate networks (ablation) | multi-task scene understanding model · Random synthetic sequences | IoU: 0.672 -> 0.764 | Improved |
| add optical flow supervision (ablation) | T/R/A/SC/S depth model · Random synthetic sequences | RMSE: 6.256 -> 6.026 | Improved |
| add smoothing loss and optical flow supervision (ablation) | T/R/A/SC depth model · Random synthetic sequences | RMSE: 6.473 -> 6.026 | Improved |
| add skip connections, smoothing loss, and optical flow supervision (ablation) | T/R/A depth model · Random synthetic sequences | RMSE: 6.826 -> 6.026 | Improved |
| add adversarial loss, skip connections, smoothing loss, and optical flow supervision (ablation) | T/R depth model · Random synthetic sequences | RMSE: 7.393 -> 6.026 | Improved |
| use temporal training and optical flow supervision instead of non-temporal training without optical flow (ablation) | depth and segmentation model · Random synthetic sequences | RMSE: 6.469 -> 6.026 | Improved |
| use the joint model rather than separate models (ablation) | scene depth and segmentation · Random synthetic sequences | depth accuracy (sigma < 1.25): 0.803 -> 0.836 | Improved |
| add the proposed method rather than DPN | Cityscapes segmentation · Cityscapes test | IoU: 66.8 -> 67.0 | Improved |
| use the proposed method rather than Reseg | CamVid segmentation · CamVid test | IoU: 58.8 -> 59.1 | Improved |
| use the proposed method rather than FDF | depth completion · Synthetic depth completion test sequences | PSNR: 46.13 -> 47.45 | Improved |
Closest papers by meaning
- Towards Scene Understanding: Unsupervised Monocular Depth Estimation With Semantic-Aware Representation (CVPR 2019 · similarity 0.88)
- Recurrent Scene Parsing With Perspective Understanding in the Loop (CVPR 2018 · similarity 0.88)
- SDC-Depth: Semantic Divide-and-Conquer Network for Monocular Depth Estimation (CVPR 2020 · similarity 0.88)
- Unleashing Semantic and Geometric Priors for 3D Scene Completion (AAAI 2026 · similarity 0.88)
- Joint Task-Recursive Learning for Semantic Segmentation and Depth Estimation (ECCV 2018 · similarity 0.87)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.86)
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.
