知识卡片

普通读书笔记卡 · 8959.d · 生成模型:GPT-6

保结构与避免复制需要权衡

理解等级:中级

较大匹配邻域可抑制局部畸形,却可能让特定叶片或区域反复出现;较小邻域增加组合自由,也可能破坏完整结构。调整邻域不能保证同时消除伪影和重复,需根据纹理尺度、参考图多样性及用途选择,并直接检查输出。

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

      Figure 28.12 shows additional image synthesis results. It is remarkable that such a simple model works so well. The results seem better than the ones shown in figure 28.9; however, note that there can still be artifacts (such as the very big leaf that appear in the output). Increasing the neighborhood size can remove those issues. Also, we can see that some elements of the

    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:摘录中关于保结构与避免复制需要权衡的定义、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:Figure 28.12 shows additional image synthesis results. It is remarkable that such a simple model works so well. The results seem better than the ones shown in figure 28.9; however, note that there can still be artifacts (such as the very big leaf that appear in the output). Increasing the neighborhood size can remove those issues. Also, we can see that some elements of the