I wanted to build my segmentation dataset for architecture like- unet/unet++. I've added a picture,so - my query is that "How do I create the true mask from an original image? " "What technology/tools used to build true mask from an original image?"
You can do it easily via using MATLAB. See for example: https://au.mathworks.com/help/images/create-binary-mask-from-grayscale-image.html
https://au.mathworks.com/matlabcentral/answers/505589-create-mask-over-image
Examples: https://au.mathworks.com/help/images/examples.html?category=building-guis-with-modular-interactive-tools&s_tid=CRUX_topnav
Muhammad Ali What if Python or others?? Matlab is not opensource & also paid.
When you talk by the true mask I understand you want to build a labeled dataset in which the mask images are ground truth for training purposes. There must be free tools on the web. However, I have in mind the imageSegmenter tool provided by Matlab in which is easy to build these true masks you need. You can create the masks using a semisupervised method called Graph Cut in which you select the background and foreground pixels before the algorithm starts. In addition, this tool has a completed manual way of creating your masks if you are looking for different and more complex shapes.
- Badges
- Science topic
- Similar topics
- Computer Science
- Images