MAL doesn't show anything at the app


I followed the notebook from the LabelBox blog and seemed worked through whole blocks.

It contains ~3000 annotations, I checked those annotations and URLs to check the predicted masks inside.

However, when I go to labeling, it doesn’t show anything.
I wonder if this feature is for paid accounts only, or if there is any other reason for that.


Hello Youjin,

I’m Ramy from Labelbox Support. I see here that you have uploaded the annotations and the screenshot shows that it seems to have gone through without throwing errors! I would suggest taking a look at this documentation and making sure that the conditions for the MAL to populate are met. Please let us know if that helps fix the issue.

Labelbox Support.

Hello @rfekry,

Thank you for a fast reply.

I checked the document and it seems my project meets all required conditions.
Here’s one of the predictions I attached.
I even made my bucket public, in the case my client cannot access it.
However, when I go check the data row with id cl4yuh2jc061i078c05e55czu in my project, I still do not see any prediction.

{'uuid': '80481b18-17e6-4cbc-a92c-92d4d8367087', 'schemaId': 'cl47nto5k1eck07angfao15gl', 'mask': {'instanceURI': '', 'colorRGB': [255, 255, 255]}, 'dataRow': {'id': 'cl4yuh2jc061i078c05e55czu'}}

For condition 2 The imported annotations are assigned to a Data Row, I labeled part of the data rows which are accepted, but I didn’t import any pre-annotations. I thought it won’t be an issue cuz some of the data rows have annotations attached to themselves. Do you think this might be the reason?

Also, I have some skipped data rows when generating predictions. I deleted them now, but I wonder if I need to follow the notebook again without any skipped data row in my project.


Thank you for that feedback! The prediction normally shows up when you hit the Start labeling button and at that point, the MAL annotations are loaded onto the asset, for that reason it may be slightly more tricky to test if the annotations really made it onto the assets.

For your second point, MAL is concerned with each row separately, meaning if you label one data row that does not affect the MAL annotations on other rows. For clarity, labeling a part of the data row would make that data row violate the third condition ( The asset has not already been labeled in the Labelbox Editor) which will not load the MAL annotations. Skipping a data row also counts as labeling the data row. Deleting and requeuing the data row should make that data row meet that third condition again and as long as all three conditions are met the MAL annotations should come back without having to reupload them i.e. rerunning the script!

Please let me know if that helps!

Labelbox Support.

1 Like

Hello @rfekry!

Thank you for your explanation! I did not know that pre-annotation only shows up on the editor.
I had a long queue of re-enqueued rows, including skipped ones and unsatisfying annotations.
After those queues is submitted, I saw new data rows with some pre-annotation.
I am not 100% sure, but it seems they are the pre-annotation from the prediction!
A lot of thanks for helping me.
I might come back again, but let me close this issue!


Great, I’m happy I was able to clarify something for you. We are always here to help! Please feel free to reach out to Labelbox Support for more detailed help!

Labelbox Support.