Visualization

[Ademord]

Hello
I am trying to use your component, and i want to visualize the 3d plot. Sample from my code:

# 3D Plot
fig = plt.figure()
ax = plt.axes(projection="3d")

# depth_image = skimage.io.imread("/host/datasets/dataset1/depth_20201030T171046.png")
image = depth_image
# filter out points that are too far
idx = np.where((image != 0) & (image <= 230))

# visualize
x_points = idx[1]
y_points = 10*image[idx]
z_points = -idx[0]
col = np.arange(30)

ax.scatter3D(x_points, y_points, z_points, c=z_points, cmap='viridis');
plt.show()

Currently what I have is a depth image, with a separated rgb.
I want to do some 3d clustering and I have been reading a lot until I found your repo for 3d connected components.

Is there a way for me to visualize the results from code you have on your main README?

[Ademord]

Also there doesnt seem to be any functional docs for using the repo on 2d images

[william-silversmith]

Hi Ademord,

Thanks for writing! I'm glad you are finding this library potentially useful.

Also there doesnt seem to be any functional docs for using the repo on 2d images

You can use 2D images just like 3D except that if you want to specify the 2D connectivity, you'll get better performance if you write connectivity=4 or connectivity=8 than with 6, 18, or 26 which are effectively 4, 8, and 8 when applied to 2D.

Currently what I have is a depth image, with a separated rgb.

cc3d only handles single channel images. To process it with cc3d, you'll need to convert the depth image from a 3-channel uint8 image (presumably) to a (u)int32 image with a single depth number per pixel.

Is there a way for me to visualize the results from code you have on your main README?

You can visualize the array in many different ways, but here is one:

pip install cloud-volume

import cc3d
from cloudvolume import view
out_labels = cc3d.connected_components(labels)
view(out_labels, segmentation=True)

That will generate a link to localhost:8080 and you can view the 3D volume by scrolling through it and reslicing different axes.

Will

[sneh-debug]

@william-silversmith how can we save the image obtained after applying cc3d?

[william-silversmith]

You can also do:

from cloudvolume.lib import save_images
out_labels = cc3d.connected_components(labels)
save_images(out_labels)

[william-silversmith]

Closing this issue due to lack of activity. Let me know if you still need help!