Applying Semantic Segmentation to own data

My question is how/where to find guidance for implementing Semantic Segmentation on our own data.
First, how to label a training set for Semantic Segmentation. How does a “trainer” generate the masks? I realize that labels should be generated for every pixel. Seems daunting; I am sure there are efficient tf/Keras ways to do that…
Thanks for your help

Hey @EduardoChicago,
I believe that applying semantic segmentation on existing datasets is out of the picture, but if that is something you want to consider, you can check this thread out.

Now, coming to your exact question, this and this thread discusses how to prepare the ground truth vectors for images in your own dataset.

Let us know if this helps.


1 Like