← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8960 · 生成模型:GPT-6

顺序纹理合成关联自回归

理解等级:高级

逐个生成像素、让新值依赖已填上下文,可按某个顺序解释为自回归条件生成。它与神经自回归共享条件分解思想,但局部经验检索、有限样本及生成顺序仍规定了不同模型。无向MRF直觉也不自动保证这些近似局部条件来自同一个精确联合分布。

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 Efros-Leung algorithm is an autoregressive model. Section 32.7 covers this class of models in more generality, and shows how to use them in conjunction with deep nets. 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

    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:摘录中关于顺序纹理合成关联自回归的定义、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 Efros-Leung algorithm is an autoregressive model. Section 32.7 covers this class of models in more generality, and shows how to use them in conjunction with deep nets. 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