@jklymak @ImportanceOfBeingErnest Thank you first.
It turns out that this confusion results from the difference between data plotting (plt.plot and the like) and image display (plt.imshow). Bellow is what I observed.

For plt.imshow(), my code is:
`import numpy as np
import matplotlib.pyplot as plt
from mxnet import image

img = image.imread("./data/dog.jpg").asnumpy()
fig, axes = plt.subplots(1, 1)
axes.imshow(img)
axes.add_patch(plt.Rectangle(xy=(200, 120), width=250, height=190,
fill=False, edgecolor='red', linewidth=3))
plt.savefig("./data/dog_boxed.jpg")`

And I got:

For plt.plot():
`x = np.linspace(0.0, 600.0, 1000)
y = 0.65 * x + 10.0

fig, axes = plt.subplots(1, 1)
axes.plot(x, y)
axes.add_patch(plt.Rectangle(xy=(100, 200), width=250, height=120,
fill=False, edgecolor='red', linewidth=3))
plt.savefig("./data/linear_boxed.png")`
And I got:

The two coordinate systems are different. For plt.plot(), matplotlib.patches.Rectangle's xy parameter is exactly right. However, for plt.imshow(), the vertical aixs is inverted.
So may some comments be added to the API reference to address this confusion?

Thanks again.

Uh oh!

Rectangle's xy parameter is wrongly explained #15401

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions