Current image annotations lost when exporting Create ML JSON

[loukrieg]

After annotating all image files in a folder, with auto save enabled, I exported a Create ML JSON file. However the annotations for the currently displayed image were not present in the exported file. To test this, I added back the annotations, moved to the next file so the annotations would be autosaved, and even clicked the SAVE button. However, the same issue occurred. On a third try, all annotations were exported successfully. Cannot reproduce the issue every time, but it has occurred more than once.

[ryouchinsa]

Thanks for writing the issue. I will do my best to reproduce the problem in my environment. Which annotation format are you using when save? PASCAL VOC xml or YOLO text format.

[ryouchinsa]

the annotations for the currently displayed image were not present in the exported file.

This means that in the Create ML JSON file, the current image exists but the annotations array is empty, correct?

{
    "annotations": [],
    "image": "current_image.jpg"
},

[loukrieg]

Yes, correct... empty annotations array (for just one image). I am using YOLO Text format.

On Dec 15, 2023, at 11:40, Ryo Kawamura @.***> wrote:

the annotations for the currently displayed image were not present in the exported file.

This means that in the Create ML JSON file, the current image exists but the annotations array is empty, correct?

{ "annotations": [], "image": "current_image.jpg" }, — Reply to this email directly, view it on GitHub https://github.com/ryouchinsa/Rectlabel-support/issues/245#issuecomment-1858164449, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7NYZNZ7OXVQBA4RBRVS2LYJR4RRAVCNFSM6AAAAABAVQT662VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNJYGE3DINBUHE. You are receiving this because you authored the thread.

[ryouchinsa]

I apologize for late reply. Using COCO validation dataset which includes 5000 images, I tried to reproduce the problem in my environment, but I could not reproduce. I really appreciate your detailed feedback.

[loukrieg]

Hi Ryo,

Thanks for trying to identify this issue.

I was exporting data yesterday in fact, and the output was perfect. Sorry I don't have a reproducible example, but I will let you know if I see the issue again.

Cheers, --Lou

On Jan 23, 2024, at 10:02, Ryo Kawamura @.***> wrote:

I apologize for late reply. Using COCO validation dataset which includes 5000 images, I tried to reproduce the problem in my environment, but I could not reproduce. I really appreciate your detailed feedback.

— Reply to this email directly, view it on GitHub https://github.com/ryouchinsa/Rectlabel-support/issues/245#issuecomment-1906237928, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA7NYZJZKSH5KUDJ7GY3KCLYP7GIBAVCNFSM6AAAAABAVQT662VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMBWGIZTOOJSHA. You are receiving this because you authored the thread.

[JohanEkenstedt]

I have also a problem to export a create ml json file. When I try to open the folder in Create ML I got the error: Empty table from specified source:

The json looks like this: [{"annotations":[{"coordinates":{"height":2579,"width":2572,"x":1547,"y":1968},"label":"Logend"}],"image":"IMG_2206.jpeg"},{"annotations":[{"coordinates":{"height":500,"width":860,"x":1429,"y":2379},"label":"Logend"}],"image":"IMG_2208.jpeg"},{"annotations":[{"coordinates":{"height":1617,"width":1673,"x":1486,"y":2204},"label":"Logend"}.....]

[ryouchinsa]

Thanks for writing the issue.

Could you check this page? "Empty table from specified data source" error in Create ML

[ryouchinsa]

Currently If you need our support to fix this problem, please let us know.