知识卡片
候选生成不是唯一选择搜索
本章以selective search概括候选区域流程,但提议可来自分割、学习网络或低成本滑窗,不应把所有这些方法等同于专名Selective Search的具体算法。比较实现时需注明提议机制、共享特征与裁剪方式。
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
- · 目标识别
原书出处 · 1 处
- · 查看原书依据
Bounding box proposal (represented as f0 in figure 50.13) can be implemented in several ways (e.g., using image segmentation [481], a neural network [404], or a window scanning approach with a low-cost classifier). The classification and bounding box refinement, f1, can be implemented by a classifier and a regression function.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第50章《目标识别》(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/c050b.xhtml)
- 结论依据:摘录中关于候选生成不是唯一选择搜索的定义、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:Bounding box proposal (represented as f0 in figure 50.13) can be implemented in several ways (e.g., using image segmentation [481], a neural network [404], or a window scanning approach with a low-cost classifier). The classification and bounding box refinement, f1, can be implemented by a classifier and a regression function.