I want to use labeled images from Image labeling in semantic segmentation,

Hello! I try to use my labeled images from Image Labeler tool, to train a network for semantic segmentation, but I have problems because those images are PNG. I have this error: Pixel label image must have 3 channels when RGB-triplet pixel label IDs are specified. My question is: Can I use these images for semantic segmentation or not? And if the answer is yes, how can I manage these images to be suitable for my porpose?

Answer 1

[yanqi liu]

yes，sir，may be modify read image function，make it to rgb，such as

trainingImages = imageDatastore('train',...
    'IncludeSubfolders',true,...
    'LabelSource','foldernames','ReadFcn',@data_preporcess);

and in read function

function data = data_preporcess(file)
data = imread(file);
if ndims(data) == 2
    data = cat(3, data, data, data);
end