← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9078.c · 生成模型:GPT-6

正则补充信息也引入偏差

理解等级:中级

数据项不充分时,正则让估计更稳定,但其输出部分来自先验而非直接观测。增大权重可能把真实速度压小或将边界抹平,应分别检查光度残差、运动边缘和可见性,而不是只追求目标函数数值更低。

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 photometric reconstruction can incorporate a regularization term penalizing fast variations on the estimated optical flow: (48.13) This problem formulation was introduced by Horn and Schunck in 1981 [218]. There are several popular regularization terms. One penalizes large velocities (slow prior): (48.14) Another penalizes variations on the optical flow (smooth prior): (48.15) The photometric loss plays an important role in unsupervised learning-based methods for optical

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第48章《光流估计》(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/c048.xhtml) - 结论依据:摘录中关于正则补充信息也引入偏差的定义、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 photometric reconstruction can incorporate a regularization term penalizing fast variations on the estimated optical flow: (48.13) This problem formulation was introduced by Horn and Schunck in 1981 [218]. There are several popular regularization terms. One penalizes large velocities (slow prior): (48.14) Another penalizes variations on the optical flow (smooth prior): (48.15) The photometric loss plays an important role in unsupervised learning-based methods for optical