PulseAugur
实时 16:57:35
English(EN) DeCoDrift: Stabilizing Decoder Coupling in Closed-Loop Foundation Segmentation

新的DeCoDrift框架稳定基础分割模型

研究人员发现基础分割模型中一种新的失败模式,称为“解码器耦合漂移”,当这些模型被迭代使用时会发生。这种漂移会导致错误累积,因为在连续迭代中模型的注意力会与目标对象失去对齐。为了解决这个问题,开发了一个名为DeCoDrift的新框架。DeCoDrift通过约束提示更新和保持解码器耦合来稳定推理时的分割过程,从而在无需重新训练或真实数据的情况下提高分割质量。 AI

影响 引入了一种提高AI模型中迭代分割任务的稳定性和准确性的方法。

排序理由 该集群包含一篇关于AI模型新研究框架的arXiv论文。

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

新的DeCoDrift框架稳定基础分割模型

报道来源 [2]

  1. arXiv cs.CV TIER_1 English(EN) · H. M. Shadman Tabib, Md. Shamsuzzoha Bayzid, M Sohel Rahman ·

    DeCoDrift: Stabilizing Decoder Coupling in Closed-Loop Foundation Segmentation

    arXiv:2605.25730v1 Announce Type: new Abstract: Foundation segmentation models such as Segment Anything Model (SAM) are now routinely used in iterative pipelines, where each predicted mask is fed back as the next prompt. This practice turns segmentation into a closed-loop dynamic…

  2. arXiv cs.CV TIER_1 English(EN) · M Sohel Rahman ·

    DeCoDrift: Stabilizing Decoder Coupling in Closed-Loop Foundation Segmentation

    Foundation segmentation models such as Segment Anything Model (SAM) are now routinely used in iterative pipelines, where each predicted mask is fed back as the next prompt. This practice turns segmentation into a closed-loop dynamical process, yet the decoder-level behavior of th…