Labelled 2D Matrix to JSON format for image segmentation deeplearning

Question

Hannes Zedel 2022-8-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1774655-labelled-2d-matrix-to-json-format-for-image-segmentation-deeplearning

回答： Shreeya 2023-9-14

Dear community,

I am trying to train a neural network for 2D panoptic image segmentation that uses JSON formatted label data as input. I look at very complex irregular shapes (metallurgical micrographs) and to avoid manual labelling, I'd like to use deterministic image analysis instead.

My thresholding-based instance segmentation method generates 2D matrices of the same dimensions as the original image and allocates each pixel to a target class by the same number, objects are identified as adjacent same numbers. I base this on the label output L of bwboundaries for each target class: [B,L]= bwboundaries()

As a snapshot, my labelled matrix/mask looks something like this (indicating 3 target classes and in this case 4 distinct objects):

[1 1 2 1 1;

1 2 2 2 1;

2 2 3 3 3]

Is there a way to convert such labels/masks, either the individual (binary) labels per target class or the complete labelled matrix, into a JSON format that I can use to train a neural network?

Thank you

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Shreeya 2023-9-14

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1774655-labelled-2d-matrix-to-json-format-for-image-segmentation-deeplearning#answer_1309626

Hi @Hannes Zedel

I understand that you have a labeled dataset for image segmentation training task and would want to convert it to JSON format. You can follow the steps outlined below to achieve it:

Begin by exporting the labels from your dataset to a .mat file or workspace. This process involves creating a ground truth object, which can be achieved using the "groundTruth" function. Detailed instructions on exporting the data and creating the ground truth object can be found in the documentation provided at the following link: https://www.mathworks.com/help/vision/ref/groundtruth.html
Once you have successfully created the ground truth object, the next step involves exporting the data to a custom JSON file format. This can be accomplished using the "exportGroundTruthToJSON" function. For a comprehensive understanding of how to utilize this function, please refer to the documentation provided at the following link: https://www.mathworks.com/help/vision/ug/export-ground-truth-object-to-custom-and-coco-json-files.html#ExportGroundTruthObjectToCustomAndCOCOJSONFilesExample-3

I hope this helps!