SPAR：面向开放词表分割的单次任意分辨率视觉 Transformer
SPAR: Single-Pass Any-Resolution ViT for Open-vocabulary Segmentation
SPAR 通过特征回归蒸馏，将滑窗教师的空间推理能力迁移到单次前向的学生模型，使 ViT 无需架构改动即可高效处理任意分辨率输入，可直接用于开放词表分割等密集预测任务的高分辨率推理。
原文依据 We distill the spatial reasoning capabilities of a finely-strided, sliding-window teacher into a single-pass student using a feature regression loss
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论基础视觉 Transformer（ViT）在需要细粒度空间理解的任务中效果有限，这源于其固定的预训练分辨率和固有的粗粒度图块级表示。这些挑战在密集预测场景中尤为突出，例如基于 ViT 的视觉语言模型进行开放词表分割时，高分辨率输入对于准确的像素级推理至关重要。现有方法通常以预训练分辨率采用滑动窗口策略处理大分辨率图像。虽然通过更细的步幅提高了精度，但计算成本显著增加。我们提出 SPAR：单次任意分辨率 ViT，一种分辨率无关的密集特征提取器，专为高效高分辨率推理设计。我们使用特征回归损失，将细步幅滑动窗口教师的空间推理能力蒸馏到单次前向的学生模型中，无需架构改动或像素级监督。应用于开放词表分割时，SPAR 将单次前向基线最高提升 10.5 平均交并比，甚至超越教师模型，证明了其在高效高分辨率推理中的有效性。代码：https://github.com/naomikombol/SPAR
英文原文摘要 1272 字符
Foundational Vision Transformers (ViTs) have limited effectiveness in tasks requiring fine-grained spatial understanding, due to their fixed pre-training resolution and inherently coarse patch-level representations. These challenges are especially pronounced in dense prediction scenarios, such as open-vocabulary segmentation with ViT-based vision-language models, where high-resolution inputs are essential for accurate pixel-level reasoning. Existing approaches typically process large-resolution images using a sliding-window strategy at the pre-training resolution. While this improves accuracy through finer strides, it comes at a significant computational cost. We introduce SPAR: Single-Pass Any-Resolution ViT, a resolution-agnostic dense feature extractor designed for efficient high-resolution inference. We distill the spatial reasoning capabilities of a finely-strided, sliding-window teacher into a single-pass student using a feature regression loss, without requiring architectural changes or pixel-level supervision. Applied to open-vocabulary segmentation, SPAR improves single-pass baselines by up to 10.5 mIoU and even surpasses the teacher, demonstrating effectiveness in efficient, high-resolution reasoning. Code: https://github.com/naomikombol/SPAR