semantic segmentation with deep learning

i have a confused between pixel label and ground truth , are there similar when using for training SegNet code . if not how can i create the pixellabels , and are 50 images enough to create pixellabels

Answer 1

[Johanna Pingel]

• Check out a few links to learn more about pixel labeling and semantic segmentation:https://www.mathworks.com/videos/semantic-segmentation-overview-1510858047780.html?s_tid=srchtitle

• You're going to want the image labeler to do the labeling: https://www.mathworks.com/help/vision/ref/imagelabeler-app.html

• The number of images really depends on the complexity of the problem. I was able to do a quick semantic segmentation with only 100 images because the images were so simple: https://blogs.mathworks.com/deep-learning/2018/11/15/sudoku-solver-image-processing-and-deep-learning/