MuViT：用于跨尺度学习的多分辨率视觉 Transformer
MuViT: Multi-Resolution Vision Transformers for Learning Across Scales in Microscopy
把不同分辨率下的 patch 嵌入统一的 world-coordinate 体系，并把 RoPE 扩展到该坐标上，使单个编码器内的注意力能同时整合大范围上下文与高分辨率细节；这套坐标化位置编码思路可直接迁移到全切片病理、遥感大幅面图像等任意多尺度视觉任务。
原文依据 MuViT embeds all patches into a shared world-coordinate system and extends rotary positional embeddings to these coordinates
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现代显微成像常规性地产生包含多个空间尺度结构的十亿像素级图像，从精细的细胞形态到更宏观的组织结构。许多分析任务需要融合这些尺度，然而大多数视觉模型只在单一分辨率下运行，或从单一视图导出多尺度特征，限制了其利用显微数据固有的多分辨率特性的能力。我们提出 MuViT，一种为融合来自同一底层图像的真实多分辨率观测而构建的 Transformer 架构。MuViT 将所有 patch 嵌入共享的世界坐标系，并将旋转位置编码扩展到这些坐标上，使注意力能够在单个编码器内将大范围上下文与高分辨率细节整合起来。在合成基准、肾脏组织病理学以及高分辨率小鼠脑显微成像上，MuViT 相较强 ViT 与 CNN 基线均取得一致提升。多分辨率 MAE 预训练进一步产生尺度一致的表征，从而增强下游任务。这些结果表明，显式的世界坐标建模为在大规模显微分析中利用多分辨率信息提供了一种简单而有力的机制。代码见 github.com/weigertlab/muvit。
英文原文摘要 1297 字符
Modern microscopy routinely produces gigapixel images that contain structures across multiple spatial scales, from fine cellular morphology to broader tissue organization. Many analysis tasks require combining these scales, yet most vision models operate at a single resolution or derive multi-scale features from one view, limiting their ability to exploit the inherently multi-resolution nature of microscopy data. We introduce MuViT, a transformer architecture built to fuse true multi-resolution observations from the same underlying image. MuViT embeds all patches into a shared world-coordinate system and extends rotary positional embeddings to these coordinates, enabling attention to integrate large-scale context with high-resolution detail within a single encoder. Across synthetic benchmarks, kidney histopathology, and high-resolution mouse-brain microscopy, MuViT delivers consistent improvements over strong ViT and CNN baselines. Multi-resolution MAE pretraining further produces scale-consistent representations that enhance downstream tasks. These results demonstrate that explicit world-coordinate modelling provides a simple yet powerful mechanism for leveraging multi-resolution information in large-scale microscopy analysis. Code is available at github.com/weigertlab/muvit.