why when I used groundTruthLabeler I got empty image (black)?

I used groundTruthLabeler and after apply pixel label end export it to folder I found it totally black as shown in attached images?how can I fix it?and groundTruthLabeler is dofferent than image segmenter app to creat groundTruth or both are doing same thing? and is it effect when I used evaluationSemanticSegementation?

[yasmin ismail]

@Image Analyst

Answer 1

[Aneela]

Hi @yasmin ismail

The black image output from the "groundTruthLabeler" can be due to the following potential reasons:

• Labeling issues: Ensure that the labels are correctly applied and visible in the app itself before exporting.
• Export settings: Make sure that the image format and settings are appropriate.
• Color Map: If you are using a colormap for visualization, ensure that it is correctly defined and applied. A mismatch in colormap settings might make the image appear black.

Ground Truth Labeler is primarily used for labelling images or videos with pixel labels, bounding boxes and scene labels while the Image Segmenter app is more focused on performing image segmentation tasks, allowing to manually segment images and create labeled data.

The "evaluateSemanticSegmentation" function relies on correctly labeled ground truth data to compute metrics like IoU (Intersection over Union), accuracy, etc. Black images without proper labels will lead to incorrect evaluation results.

For more information on Ground Truth Labeler and Image Segmenter app, refer to the following MathWorks documentation links respectively:

https://www.mathworks.com/help/driving/ref/groundtruthlabeler-app.html

https://www.mathworks.com/help/images/ref/imagesegmenter-app.html