Research map / 3D vision / Depth and stereo
EventHub: Data Factory for Generalizable Event-Based Stereo Networks without Active Sensors
Luca Bartolomei, Fabio Tosi, Matteo Poggi, Stefano Mattoccia, Guillermo Gallego
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 DepthAnythingV2 depth regularizer to SVRaster (ablation) | SVRaster NVS · ScanNet++ subset | depth MAE (cm): 8.99 -> 6.61 | Improved |
| add MASt3R regularizer on top of all other regularizers (ablation) | SVRaster NVS · ScanNet++ subset | depth MAE (cm): 6.61 -> 38.71 | Hurt |
| add normal-consistency regularizers (with DAv2) (ablation) | SVRaster NVS · ScanNet++ subset | depth MAE (cm): 7.15 -> 6.57 | Improved |
| add asc and sparse regularizers (ablation) | SVRaster NVS · ScanNet++ subset | depth MAE (cm): 6.58 -> 6.57 | No clear effect |
| halve the normal regularizer weights (ablation) | SVRaster NVS · ScanNet++ subset | depth MAE (cm): 6.57 -> 6.38 | Improved |
| replace ambient-occlusion confidence with voxel-size confidence (ablation) | EventHub NVS labels · ScanNet++ subset (threshold 0.85 vs 0.45) | depth MAE (cm): 6.20 -> 6.03 | Improved |
| use SVRaster instead of Depth Anything v3 as rendering engine (ablation) | EventHub MIX 3 SE-CFF · DSEC val | MAE (px): 3.18 -> 1.01 | Improved |
| add ScanNet++ data to NeRF-Stereo data (MIX 1 to MIX 2) (ablation) | E-FoundationStereo · DSEC val | MAE (px): 1.39 -> 1.04 | Improved |
| add NeRF-Stereo and ScanNet++ to DSEC distilled data (MIX 3 to MIX 4) (ablation) | SE-CFF · DSEC val | MAE (px): 1.01 -> 1.07 | Hurt |
| add NeRF-Stereo and ScanNet++ to DSEC distilled data (MIX 3 to MIX 4) (ablation) | E-FoundationStereo · M3ED Day | MAE (px): 1.49 -> 0.98 | Improved |
| extend proxy-event synthetic data with TartanAir (ablation) | EV-SceneFlow E-FoundationStereo · DSEC val | MAE (px): 3.10 -> 1.76 | Improved |
| train with MIX 4 proxy labels instead of DSEC LiDAR labels | E-FoundationStereo · MVSEC Day | MAE (px): 6.04 -> 1.33 | Improved |
Closest papers by meaning
- Temporal Event Stereo via Joint Learning with Stereoscopic Flow (ECCV 2024 · similarity 0.90)
- Learning an Event Sequence Embedding for Dense Event-Based Deep Stereo (ICCV 2019 · similarity 0.89)
- Deep Event Stereo Leveraged by Event-to-Image Translation (AAAI 2021 · similarity 0.89)
- Active Event-based Stereo Vision (CVPR 2025 · similarity 0.88)
- Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain (NeurIPS 2024 · similarity 0.87)
- DERD-Net: Learning Depth from Event-based Ray Densities (NeurIPS 2025 · 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.
