知识卡片
学习抓取模型仍受训练分布限制
图像与深度可通过学习模型融合并生成抓取候选,减少手写处理环节。其优势需在目标物体和环境验证,不能从“数据驱动”推出始终比手工管线更好或省去碰撞与执行确认。
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
- · 操纵
原书出处 · 1 处
- · 查看原书依据
With the recent advent of (deep) machine learning and the ability of neural networks (chapter 10) to approximate complex functions, it is also possible to replace parts or all of the algorithmic steps shown in figure 14.1 with a convolutional neural network trained by deep learning. While data intensive, such an approach can seamlessly merge image and depth data and adapt to application-specific data better
参考来源
- 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第14章《操纵》(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_14.xhtml)
- 结论依据:摘录中关于学习抓取模型仍受训练分布限制的定义、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 recent advent of (deep) machine learning and the ability of neural networks (chapter 10) to approximate complex functions, it is also possible to replace parts or all of the algorithmic steps shown in figure 14.1 with a convolutional neural network trained by deep learning. While data intensive, such an approach can seamlessly merge image and depth data and adapt to application-specific data better