Research map / 3D vision / Depth and stereo
EventVGGT: Exploring Cross-Modal Distillation for Consistent Event-based Depth Estimation
Yinrui Ren, Jinjing Zhu, Kanghao Chen, Zhuoxiao Li, Jing OU, Zidong Cao, Tongyan Hua, Peilun Shi, Yingchun Fu, Wufan Zhao, Hui Xiong
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 CMFM loss (ablation) | EventVGGT baseline · EventScape (Town 1 training) | absolute mean depth error 30m: 1.26 -> 1.13 | Improved |
| add STFD on top of CMFM (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.13 -> 1.10 | Improved |
| add TCD on top of CMFM and STFD (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.10 -> 1.06 | Improved |
| add TCD on top of CMFM only (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.13 -> 1.12 | Improved |
| increase input sequence length from 1 to 24 frames (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.50 -> 1.26 | Improved |
| increase sequence length from 1 to 8 frames (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.50 -> 1.31 | Improved |
| lower CMFM mixture rate from 100% to 25% (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.26 -> 1.13 | Improved |
| lower CMFM mixture rate from 25% to 10% (ablation) | EventVGGT · EventScape (Town 1 training) | absolute mean depth error 30m: 1.13 -> 1.16 | Hurt |
| train on Towns 1-3 instead of Town 1 only (ablation) | EventVGGT · DENSE (zero-shot) | absolute mean depth error 30m: 1.50 -> 1.33 | Improved |
| train on Towns 1-3 instead of Town 1 only (ablation) | EventVGGT · EventScape | absolute mean depth error 30m: 1.09 -> 1.06 | No clear effect |
| distill VGGT into event student instead of applying VGGT directly to events | EventVGGT · MVSEC Night1 (zero-shot) | absolute mean depth error 30m: 3.44 -> 2.61 | Improved |
| event-only EventVGGT vs EventDAM | EventVGGT · EventScape | absolute mean depth error 30m: 2.30 -> 1.06 | Improved |
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.
- Depth AnyEvent — Depth AnyEvent: A Cross-Modal Distillation Paradigm for Event-Based Monocular Depth Estimation (ICCV 2025) · better on 4 of 4 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
- Depth AnyEvent: A Cross-Modal Distillation Paradigm for Event-Based Monocular Depth Estimation (ICCV 2025 · similarity 0.94)
- Depth Any Event Stream: Enhancing Event-based Monocular Depth Estimation via Dense-to-Sparse Distillation (ICCV 2025 · similarity 0.93)
- Stereo Depth From Events Cameras: Concentrate and Focus on the Future (CVPR 2022 · similarity 0.91)
- Event-Driven Dynamic Scene Depth Completion (NeurIPS 2025 · similarity 0.90)
- Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain (NeurIPS 2024 · similarity 0.89)
- DERD-Net: Learning Depth from Event-based Ray Densities (NeurIPS 2025 · similarity 0.88)
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.
