Scaling Dense Event-Stream Pretraining from：它和通用视觉自监督的关系在于：event camera 模态较专，但 VFM 蒸馏到异构视觉流的结构对齐思路可迁移
中相关；详见方法、贡献和实验边界。
Scaling Dense Event-Stream Pretraining from Visual Foundation Models arXiv + CVPR 2026 Day 1 补录 原文链接
编号2603.03969优先级扫读类别CVPR 2026 Day 1 补录会议arXiv + CVPR 2026 Day 1 补录方法event camera 模态较专，但 VFM 蒸馏到异构视觉流的结构对齐思路可迁移来源arXiv / OpenReview
先说结论。它和通用视觉自监督的关系在于：event camera 模态较专，但 VFM 蒸馏到异构视觉流的结构对齐思路可迁移。 中相关；详见方法、贡献和实验边界。
Figureure 10 · The qualitative comparisons among different event-based semantic segmeFigure 10. The qualitative comparisons among different event-based semantic segmention approaches on the test set of DSEC-Semantic. Best viewed in color.这张图/表用于判断 Scaling Dense Event-Stream Pretraining from 的实验收益来自哪里。重点看替换、消融或跨模型设置下趋势是否一致，而不是只看单个最高分。Comparison of dense event features of 不同的蒸馏策略Comparison of dense event features of 不同的蒸馏策略. 随着深入分辨率的提升，事件表征会出现不同程度的退化： PCA maps become less localized Figure 4. Comparison of dense event features under different distillation strategies. All features are produced by a DINOv3-ViT-B and the similarity maps (marked with a red point) become noisier. 所有的特征由DINOv3-ViTB模型和不参与蒸馏的测试样本输出. From up to down: (a) model. Left to right: as spatial resolution increases, event representations degrade to varying degrees. PCA maps become less localized, image 表征; (b)\~(e)表示由不同蒸馏策略得到的事件表征, (a) patch-level 蒸馏, (b) superpixel-level 蒸馏, (c) patch-level 蒸馏 w/ Event Activation Mask, (d) and similarity maps (anchored at the red dot) become noisier. Top to bottom: (a) image features; (b) patch-level distillation; (c) superpixellevel distillation; (d) patch-level distillation + event activation mask; (e) patch-level distillation + event activation mask + structure-aware regularization (Our method). A more detailed feature analysis is provided in the supplementary materials.这张图/表用于判断 Scaling Dense Event-Stream Pretraining from 的实验收益来自哪里。重点看替换、消融或跨模型设置下趋势是否一致，而不是只看单个最高分。
核心问题
它和通用视觉自监督的关系在于：event camera 模态较专，但 VFM 蒸馏到异构视觉流的结构对齐思路可迁移。
