知识卡片
密集预测保留空间结构
在每个位置输出类别向量可以形成语义分割图,而全局汇聚后则可输出整图类别。CNN可以构建两类任务,是否保留输入大小由步长、padding、上下采样与输出头决定,不能说所有CNN都必然输出同形状图像。
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
- · 卷积神经网络
原书出处 · 1 处
- · 查看原书依据
Each patch is processed with a classifier module, which is a neural net. Essentially, this neural net scans across the patches in the input and classifies each. The output is a label for each patch in the input image. If we rearrange these predictions back into the shape of the input image and color code them, we get the below input-output mapping (figure 24.2):
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第24章《卷积神经网络》(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/c024b.xhtml)
- 结论依据:摘录中关于密集预测保留空间结构的定义、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:Each patch is processed with a classifier module, which is a neural net. Essentially, this neural net scans across the patches in the input and classifies each. The output is a label for each patch in the input image. If we rearrange these predictions back into the shape of the input image and color code them, we get the below input-output mapping (figure 24.2):