axes

一： fig.add_axes 画图中图 fig = plt.figure() x = np.arange(1, 9, 1) y = np.linspace(1, 10, 8) left, bottom, width, height = 0.1, 0.1, 0.8, 0.8 ax1 = fig.add_axes([left, bottom, width, height]) ax1.plot(x, y, 'r') ax1.set_xlabel('x') ax1.set_ylabel('y') ax1.set_title('title') left, bottom, width, height = 0.2, 0.6, 0.25, 0.25 ax2 = fig.add_axes([left, bottom, width, height]) ax2.plot(y, x, 'b') ax2.set_xlabel('x') ax2.set_ylabel('y') ax2.set_title('title inside 1') left, bottom, width, height = 0.6, 0.2, 0.25, 0.25 ax3 = fig.add_axes([left, bottom, width, height]) ax3.plot(y, x, 'g') ax3.set_xlabel(

目录 im2col简介 im2col.cpp源码 小结 参考 @(Caffe源码-im2col操作) im2col简介 caffe的卷积操作中使用im2col来加速，im2col将卷积核中的每个点在图像上的对应点全都提取出来按行排列，得到一个矩阵，这样就将卷积操作转化为矩阵进行操作。 如上图所示的，假设输入图像的形状为 channels=1, height=width=5 ，并且 pad_w=pad_h=1, kernel_h=kernel_w=3, stride_h=stride_w=2, dilation_w=dilation_h=1 。左侧图中蓝色为padding补充的边界，值均为0，绿色为实际图像的数据。其中卷积核中 \(k_{00}\) 位置在整个卷积操作中共计算了 output_h*output_w=9 次，每次的位置在左侧图中用黑色实心圆标注出来。而im2col操作即是将卷积核上的每个点的这些对应位置上的值都提取出来，按照右侧黄色方格的形式存放起来。这样卷积操作可简单地通过将卷积核（中间的红色方格）展成一个向量，然后与右侧的黄色方格矩阵中的每一列点乘来实现。更详细的说明可查看后面列出来的参考博客。 与im2col对应的是col2im操作，即是将矩阵还原成卷积前的图像的形状，不过caffe代码中的 col2im_cpu() 函数还稍微有些改动。 im2col.cpp源码

As I mentioned in my last post, I have begun a project using the Flare data visualization toolkit , and I intend to document things I learn along the way. One of the first things I investigated, after familiarizing myself with the basics was how customizable the grids and axes for visualizations could be. It took me longer than I expected to figure things out, but now that I have I would say the grids and axes in Flare are highly customizable, and should be capable of accommodating just about any needs you may have. This post is mostly code and images, but in case you want to jump around here