一篇新论文评估了 Segment Anything Model 3 (SAM 3) 在遥感任务中的能力，发现它虽然避免了过拟合并在分割方面表现良好，但在亚像素分辨率和语义盲点方面存在困难。该研究引入了一种方法来使 SAM 3 适应零样本分类，并通过分离文本和视觉提示来分析其多模态解码器。研究表明，视觉提示使模型与复杂的地缘空间几何对齐，但文本提示引入了不匹配的地面语义偏差，阻碍了性能。 AI
影响 强调了在遥感等专业应用中，对基础模型进行领域特定微调的必要性。
排序理由 学术论文，评估现有模型在特定领域的能力。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2607.09583v1 Announce Type: new Abstract: The deployment of large-scale foundation models, such as the Segment Anything Model 3 (SAM 3), promises a transition toward open-vocabulary, training-free computer vision. However, their capacity to generalize out-of-distribution to…
The deployment of large-scale foundation models, such as the Segment Anything Model 3 (SAM 3), promises a transition toward open-vocabulary, training-free computer vision. However, their capacity to generalize out-of-distribution to the complex, top-down geometric structures of E…