Research map / 3D vision / Depth and stereo
Physics-Based Differentiable Depth Sensor Simulation
Benjamin Planche, Rajat Vikram Singh
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 3D clutter to the virtual scenes (ablation) | DDS (untrained) + DR · Cropped LineMOD | accuracy: 50.0 -> 89.6 | Improved |
| add 3D clutter to the virtual scenes (ablation) | DDS (|Phi|=2,535) + DR · Cropped LineMOD | accuracy: 54.3 -> 93.0 | Improved |
| add 3D clutter to the virtual scenes (ablation) | Basic · Cropped LineMOD | accuracy: 21.3 -> 46.8 | Improved |
| add online domain randomization (DR) (ablation) | DDS (untrained) · Cropped LineMOD | accuracy: 69.7 -> 89.6 | Improved |
| add online domain randomization (DR) (ablation) | Basic · Cropped LineMOD | accuracy: 46.8 -> 70.7 | Improved |
| optimize 4 simulation parameters unsupervised (ablation) | DDS · Cropped LineMOD | accuracy: 69.7 -> 81.2 | Improved |
| add two trainable post-processing convolutions (2,535 parameters) (ablation) | DDS (|Phi|=4) · Cropped LineMOD | accuracy: 81.2 -> 85.5 | Improved |
| add two trainable post-processing convolutions with DR (ablation) | DDS (|Phi|=4) + DR · Cropped LineMOD | accuracy: 90.5 -> 93.0 | Improved |
| add real annotated training data to DDS (ablation) | DDS (|Phi|=2,535) + DR · Cropped LineMOD | accuracy: 93.0 -> 97.8 | Improved |
| train on DDS data instead of real data (ablation) | T trained on real data · Cropped LineMOD | rotation error: 35.0 -> 31.3 | Improved |
| train on DDS data instead of real data (ablation) | T trained on real data · Cropped LineMOD | accuracy: 95.4 -> 93.0 | Hurt |
| train DDS on scene 3 real scans (supervised) (ablation) | DDS (untrained) · 2D-3D-S segmentation | pixel accuracy: 62.9 -> 69.8 | Improved |
Closest papers by meaning
- HyperDepth: Learning Depth From Structured Light Without Matching (CVPR 2016 · similarity 0.87)
- Domain Randomization-Enhanced Depth Simulation and Restoration for Perceiving and Grasping Specular and Transparent Objects (ECCV 2022 · similarity 0.86)
- ∂DIBR: Differentiable Depth Image-based Rendering for Fast Novel View Synthesis (ECCV 2026 · similarity 0.86)
- Efficient and Differentiable Shadow Computation for Inverse Problems (ICCV 2021 · similarity 0.86)
- Configurable 3D Scene Synthesis and 2D Image Rendering with Per-pixel Ground Truth Using Stochastic Grammars (IJCV 2018 · similarity 0.86)
- Differentiable rendering with perturbed optimizers (NeurIPS 2021 · 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.
