Pixel label image has scalar pixel label IDs instead of RGB-triplet pixel label IDs

Hi, I am trying to follow this semantic segmentation using deep learning tutorial with my own dataset.

https://www.mathworks.com/help/vision/ug/semantic-segmentation-using-deep-learning.html

I am running into this error when I am trying to train the network.

I am not sure if this error has to do with how we are labeling the images. I am labeling all my images with imageLabeler() but I noticed we get all black images in the PixelLabelData folder after we export the label data. In the camvid dataset example, the label data had the color overlay of each different label. Here is one of the pictures I labeled and the black image I get versus the camvid one with color overlay I am talking about.

How can fix this error so I can continue training the network? Any help will be greatly appreciated. Thanks in advance!

[Nipuna]

Hi did you get a reasonable answer for this problem? I'm trying the original example and faced the same issue.

Answer 1

[Mahesh Taparia]

Hi

The labeled images have label ID (for Eg: 1,2,3 etc. upto number of classes). So it look black as label ID value is small. To visualize the labels on input image, you can use labeloverlay function. So your 1st image is label image, if you read it in MATLAB, you will get the respective pixel id of each class.These images are used as label images while doing training. Hope it will help!