INSID3：基于 DINOv3 的免训练上下文分割
INSID3: Training-Free In-Context Segmentation with DINOv3
利用冻结的 DINOv3 密集自监督特征直接进行语义匹配和分割，无需微调或辅助模型，为少样本分割提供了极简且强泛化的新范式。
原文依据 We introduce INSID3, a training-free approach that segments concepts at varying granularities only from frozen DINOv3 features
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论上下文分割（ICS）旨在给定一个标注视觉示例的情况下分割任意概念，例如物体、部件或个性化实例。现有工作依赖于（i）微调视觉基础模型（VFM），这提高了域内结果但损害了泛化能力，或（ii）组合多个冻结的 VFM，这保持了泛化性但导致架构复杂性和固定的分割粒度。我们从极简主义视角重新审视 ICS，并提问：一个单一的自监督骨干网络能否在没有任何监督或辅助模型的情况下同时支持语义匹配和分割？我们表明，来自 DINOv3 的放大密集自监督特征表现出强大的空间结构和语义对应关系。我们引入 INSID3，一种免训练方法，仅从冻结的 DINOv3 特征中，在给定上下文示例的情况下，以不同粒度分割概念。INSID3 在单样本语义、部件和个性化分割上取得了最优结果，比先前工作高出 +7.5% mIoU，同时使用的参数少 3 倍，且无需任何掩码或类别级监督。
英文原文摘要 1130 字符
In-context segmentation (ICS) aims to segment arbitrary concepts, e.g., objects, parts, or personalized instances, given one annotated visual examples. Existing work relies on (i) fine-tuning vision foundation models (VFMs), which improves in-domain results but harms generalization, or (ii) combines multiple frozen VFMs, which preserves generalization but yields architectural complexity and fixed segmentation granularities. We revisit ICS from a minimalist perspective and ask: Can a single self-supervised backbone support both semantic matching and segmentation, without any supervision or auxiliary models? We show that scaled-up dense self-supervised features from DINOv3 exhibit strong spatial structure and semantic correspondence. We introduce INSID3, a training-free approach that segments concepts at varying granularities only from frozen DINOv3 features, given an in-context example. INSID3 achieves state-of-the-art results across one-shot semantic, part, and personalized segmentation, outperforming previous work by +7.5 % mIoU, while using 3x fewer parameters and without any mask or category-level supervision.