search

NVlabs / ssn_superpixels

Superpixel Sampling Networks (ECCV2018)
https://varunjampani.github.io/ssn/
Other
350 stars 56 forks source link

superpixel results #14

Closed zicair closed 5 years ago

zicair commented 5 years ago

Thank you for the release of project.

But I have a question, using pre-trained model to run directly, the result contains 'new_spix_indices' and '_bdry.jpg' in the code. How can I get results like this:

image

Looking forward to your reply, thank you!

varunjampani commented 5 years ago

Can you explain more clearly? What do you mean 'result'? Do you mean that 'visualization'?

zicair commented 5 years ago

I am sorry，I don't know much about superpixels. All I know about superpixels is that you get the effect where the value of each superpixel block is the average of the pixels in that area. For example，left image is raw image and right one is the visualization result.

image

The screenshot above was taken from https://github.com/PSMM/SLIC-Superpixels

varunjampani commented 5 years ago

Superpixels give 'segments/clusters' (over-segmentation) of pixels an image. 'new_spix_indices' give unique number to pixels belonging to a superpixel cluster. Using those new_spix_indices, one could average the pixel values in each superpixel cluster to generate an average image.

varunjampani commented 5 years ago

This could be useful to learn about superpixel visualization: https://www.pyimagesearch.com/2017/06/26/labeling-superpixel-colorfulness-opencv-python/

zicair commented 5 years ago

I get it, thx！