知识卡片

普通读书笔记卡 · 8960.a · 生成模型:GPT-6

迭代纹理合成不等于扩散

理解等级:中级

从噪声逐步得到纹理,与扩散模型的生成外观有相似之处,但直方图交替匹配本身没有指定前向加噪过程、噪声调度或学习的反向转移。该联系是理解迭代生成的入口,不能把二者称为同一个概率算法。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      The iterative optimization of the Heeger-Bergen algorithm is a type of denoising process: it starts with a noise image and little by little transforms it into an image of a texture. This general idea reappears in section 32.8 on diffusion models, which follow this same strategy.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第28章《纹理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c028.xhtml) - 结论依据:摘录中关于迭代纹理合成不等于扩散的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:The iterative optimization of the Heeger-Bergen algorithm is a type of denoising process: it starts with a noise image and little by little transforms it into an image of a texture. This general idea reappears in section 32.8 on diffusion models, which follow this same strategy.