Research map / 3D vision / Depth and stereo
Depth Any Event Stream: Enhancing Event-based Monocular Depth Estimation via Dense-to-Sparse Distillation
Jinjing Zhu, Tianbo Pan, Zidong Cao, Yexin Liu, James T. Kwok, 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 SFM loss to the baseline (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (1.35, 3.34, 4.92) -> (0.62, 1.75, 2.73) | Improved |
| add SFD loss to the baseline (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (1.35, 3.34, 4.92) -> (0.70, 1.79, 2.76) | Improved |
| add SCM to the baseline (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (1.35, 3.34, 4.92) -> (0.67, 1.78, 2.72) | Improved |
| add SCM on top of SFM + SFD (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.60, 1.64, 2.57) -> (0.59, 1.60, 2.47) | Improved |
| use sparsity-guided feature mixing instead of random mixing (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.76, 2.15, 3.50) -> (0.62, 1.75, 2.73) | Improved |
| use sparsity-guided feature mixing instead of no mixing (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.65, 1.92, 2.99) -> (0.62, 1.75, 2.73) | Improved |
| add sparsity guidance to feature distillation (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.79, 2.01, 3.00) -> (0.70, 1.79, 2.76) | Improved |
| use LoRA encoder with fixed decoder instead of fine-tuning all layers (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.80, 2.27, 3.33) -> (0.60, 1.64, 2.57) | Improved |
| keep decoder fixed instead of fine-tuning it with LoRA encoder (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 10m/20m/30m: (0.69, 1.87, 2.88) -> (0.60, 1.64, 2.57) | Improved |
| fine-tune DAM with LoRA instead of fixed model (ablation) | EventDAM-S · EventScape ablation (DAM ViT-S) | depth error 30m: 4.92 -> 2.57 | Improved |
| distill DAM into event model (EventDAM-S) instead of using DAM-S on event input | DAM-S · MVSEC Night1 zero-shot | depth error 30m: 4.25 -> 3.36 | Improved |
| use event-only EventDAM-S instead of image-input DAM-S | EventDAM-S · MVSEC Night1 zero-shot | depth error 30m: 3.06 -> 3.36 | Hurt |
Closest papers by meaning
- Depth AnyEvent: A Cross-Modal Distillation Paradigm for Event-Based Monocular Depth Estimation (ICCV 2025 · similarity 0.94)
- EventVGGT: Exploring Cross-Modal Distillation for Consistent Event-based Depth Estimation (ECCV 2026 · similarity 0.93)
- Event-Driven Dynamic Scene Depth Completion (NeurIPS 2025 · similarity 0.91)
- Distil-E2D: Distilling Image-to-Depth Priors for Event-Based Monocular Depth Estimation (NeurIPS 2025 · similarity 0.91)
- DERD-Net: Learning Depth from Event-based Ray Densities (NeurIPS 2025 · similarity 0.89)
- Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain (NeurIPS 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.
