知识卡片
降噪不能抹掉任务所需证据
预处理把图像变得便于后续计算,同时可能去掉小结构或弱纹理。端到端学习可把部分特征提取整合进任务训练,但并非每个网络都有显式形态学或固定阈值;仍需按最终识别与几何目标判断信息是否保留。
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
- · 视觉
原书出处 · 1 处
- · 查看原书依据
3. There is a trade-off between making the data stream more tractable and preserving actual information. As computers and algorithms, in particular machine learning, become more powerful, modern vision systems often blend preprocessing and actual image understanding into a single pipeline.
参考来源
- 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第08章《视觉》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_8.xhtml)
- 结论依据:摘录中关于降噪不能抹掉任务所需证据的定义、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:3. There is a trade-off between making the data stream more tractable and preserving actual information. As computers and algorithms, in particular machine learning, become more powerful, modern vision systems often blend preprocessing and actual image understanding into a single pipeline.