Python Tensorflow：如何在 Tensorboard 中显示自定义图像（例如 Matplotlib 图）
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/38543850/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
Tensorflow: How to Display Custom Images in Tensorboard (e.g. Matplotlib Plots)
提问by RobR
The Image Dashboardsection of the Tensorboard ReadMe says:
Tensorboard 自述文件的图像仪表板部分说：
Since the image dashboard supports arbitrary pngs, you can use this to embed custom visualizations (e.g. matplotlib scatterplots) into TensorBoard.
由于图像仪表板支持任意 png，您可以使用它来将自定义可视化（例如 matplotlib 散点图）嵌入到 TensorBoard 中。
I see how a pyplot image could be written to file, read back in as a tensor, and then used with tf.image_summary() to write it to TensorBoard, but this statement from the readme suggests there is a more direct way. Is there? If so, is there any further documentation and/or examples of how to do this efficiently?
我看到如何将 pyplot 图像写入文件，作为张量读回，然后与 tf.image_summary() 一起使用以将其写入 TensorBoard，但自述文件中的此语句表明有一种更直接的方法。在那儿？如果是这样，是否有任何进一步的文档和/或示例说明如何有效地做到这一点？
回答by Andrzej Pronobis
It is quite easy to do if you have the image in a memory buffer. Below, I show an example, where a pyplot is saved to a buffer and then converted to a TF image representation which is then sent to an image summary.
如果您将图像放在内存缓冲区中，这很容易做到。下面，我展示了一个示例，其中将 pyplot 保存到缓冲区，然后转换为 TF 图像表示，然后将其发送到图像摘要。
import io
import matplotlib.pyplot as plt
import tensorflow as tf
def gen_plot():
"""Create a pyplot plot and save to buffer."""
plt.figure()
plt.plot([1, 2])
plt.title("test")
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
return buf
# Prepare the plot
plot_buf = gen_plot()
# Convert PNG buffer to TF image
image = tf.image.decode_png(plot_buf.getvalue(), channels=4)
# Add the batch dimension
image = tf.expand_dims(image, 0)
# Add image summary
summary_op = tf.summary.image("plot", image)
# Session
with tf.Session() as sess:
# Run
summary = sess.run(summary_op)
# Write summary
writer = tf.train.SummaryWriter('./logs')
writer.add_summary(summary)
writer.close()
This gives the following TensorBoard visualization:
这给出了以下 TensorBoard 可视化：
回答by y.selivonchyk
Next script does not use intermediate RGB/PNG encoding. It also fixes the issue with additional operation construction during execution, single summary is reused.
Next 脚本不使用中间 RGB/PNG 编码。它还解决了在执行期间附加操作构造的问题，重复使用单个摘要。
Size of the figure is expected to remain the same during execution
预计在执行期间图形的大小将保持不变
Solution that works:
有效的解决方案：
import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np
def get_figure():
fig = plt.figure(num=0, figsize=(6, 4), dpi=300)
fig.clf()
return fig
def fig2rgb_array(fig, expand=True):
fig.canvas.draw()
buf = fig.canvas.tostring_rgb()
ncols, nrows = fig.canvas.get_width_height()
shape = (nrows, ncols, 3) if not expand else (1, nrows, ncols, 3)
return np.fromstring(buf, dtype=np.uint8).reshape(shape)
def figure_to_summary(fig):
image = fig2rgb_array(fig)
summary_writer.add_summary(
vis_summary.eval(feed_dict={vis_placeholder: image}))
if __name__ == '__main__':
# construct graph
x = tf.Variable(initial_value=tf.random_uniform((2, 10)))
inc = x.assign(x + 1)
# construct summary
fig = get_figure()
vis_placeholder = tf.placeholder(tf.uint8, fig2rgb_array(fig).shape)
vis_summary = tf.summary.image('custom', vis_placeholder)
with tf.Session() as sess:
tf.global_variables_initializer().run()
summary_writer = tf.summary.FileWriter('./tmp', sess.graph)
for i in range(100):
# execute step
_, values = sess.run([inc, x])
# draw on the plot
fig = get_figure()
plt.subplot('111').scatter(values[0], values[1])
# save the summary
figure_to_summary(fig)
回答by cheind
A bit late with my answer. With tf-matplotliba simple scatter plot boils down to:
我的回答有点晚了。使用tf-matplotlib一个简单的散点图可以归结为：
import tensorflow as tf
import numpy as np
import tfmpl
@tfmpl.figure_tensor
def draw_scatter(scaled, colors):
'''Draw scatter plots. One for each color.'''
figs = tfmpl.create_figures(len(colors), figsize=(4,4))
for idx, f in enumerate(figs):
ax = f.add_subplot(111)
ax.axis('off')
ax.scatter(scaled[:, 0], scaled[:, 1], c=colors[idx])
f.tight_layout()
return figs
with tf.Session(graph=tf.Graph()) as sess:
# A point cloud that can be scaled by the user
points = tf.constant(
np.random.normal(loc=0.0, scale=1.0, size=(100, 2)).astype(np.float32)
)
scale = tf.placeholder(tf.float32)
scaled = points*scale
# Note, `scaled` above is a tensor. Its being passed `draw_scatter` below.
# However, when `draw_scatter` is invoked, the tensor will be evaluated and a
# numpy array representing its content is provided.
image_tensor = draw_scatter(scaled, ['r', 'g'])
image_summary = tf.summary.image('scatter', image_tensor)
all_summaries = tf.summary.merge_all()
writer = tf.summary.FileWriter('log', sess.graph)
summary = sess.run(all_summaries, feed_dict={scale: 2.})
writer.add_summary(summary, global_step=0)
When executed, this results in the following plot inside Tensorboard
执行时，这会在 Tensorboard 中生成以下图
Note that tf-matplotlibtakes care about evaluating any tensor inputs, avoids pyplotthreading issues and supports blitting for runtime critical plotting.
请注意，tf-matplotlib负责评估任何张量输入，避免pyplot线程问题并支持用于运行时关键绘图的 blitting。
回答by subjord
Finally there is some official documentationabout "Logging arbitrary image data" with an example of matplotlib created images.
最后有一些关于“记录任意图像数据”的官方文档，其中包含一个 matplotlib 创建的图像示例。
回答by fr_andres
This intends to complete Andrzej Pronobis' answer. Following closely his nice post, I set up this minimal working example:
这打算完成 Andrzej Pronobis 的回答。密切关注他的好帖子，我设置了这个最小的工作示例：
plt.figure()
plt.plot([1, 2])
plt.title("test")
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
image = tf.image.decode_png(buf.getvalue(), channels=4)
image = tf.expand_dims(image, 0)
summary = tf.summary.image("test", image, max_outputs=1)
writer.add_summary(summary, step)
Where writer is an instance of tf.summary.FileWriter. This gave me the following error: AttributeError: 'Tensor' object has no attribute 'value'For which this github posthad the solution: the summary has to be evaluated (converted into a string) before being added to the writer. So the working code for me remained as follows (simply add the .eval() call in the last line):
其中 writer 是tf.summary.FileWriter 的一个实例。这给了我以下错误： AttributeError: 'Tensor' object has no attribute 'value'对于这个 github post有解决方案：在将摘要添加到 writer 之前，必须对其进行评估（转换为字符串）。所以我的工作代码如下（只需在最后一行添加 .eval() 调用）：
plt.figure()
plt.plot([1, 2])
plt.title("test")
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
image = tf.image.decode_png(buf.getvalue(), channels=4)
image = tf.expand_dims(image, 0)
summary = tf.summary.image("test", image, max_outputs=1)
writer.add_summary(summary.eval(), step)
This could be short enough to be a comment on his answer, but these can be easily overlooked (and I may be doing something else different too), so here it is, hope it helps!
这可能足够短，可以作为对他的回答的评论，但这些很容易被忽视（我也可能在做其他不同的事情），所以就在这里，希望它有所帮助！
Cheers,
Andres
干杯，
安德烈斯
