知识卡片
可诊断失败未必容易修复
显式管线使错误可追溯到边缘、分割或几何假设,但知道哪一步失败不一定能在所有新场景修好。可解释性与泛化能力是不同性质,设计时应分别检验,不能将一者当作另一者的保证。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 重访简单视觉系统
原书出处 · 1 处
- · 查看原书依据
With the hand-crafted model presented in chapter 2, we had a comprehensive understanding of the mechanisms, assumptions, and limitations of our system. Another interesting advantage of the hand-crafted model is that it required no learning and no training data. Once the model was written down it was ready to work. It worked really well when the conditions were met, but it was fragile and unreliable
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第55章《重访简单视觉系统》(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/c055b.xhtml)
- 结论依据:摘录中关于可诊断失败未必容易修复的定义、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:With the hand-crafted model presented in chapter 2, we had a comprehensive understanding of the mechanisms, assumptions, and limitations of our system. Another interesting advantage of the hand-crafted model is that it required no learning and no training data. Once the model was written down it was ready to work. It worked really well when the conditions were met, but it was fragile and unreliable