Converting Labelme Annotations to Label Studio pre-annotated annotations

Hello all!

I and my team need to migrate to Label Studio.

Previously, we used Labelme, which works well for text segmentation. However, we needed to create a custom task for text segmentation + text transcription. It is quite similar to the OCR setup label studio has, really.

We uploaded labelme's annotation results (.json) and their respective images (.png) to 2 separate folders in a s3 bucket.

The issue is, while trying to convert Labelme's annotation format into Label Studio's, we fail to load the image available in the s3 bucket to Label Studio.

Issues:
1. Label studio's cloud storage only takes .json, so the images must be saved on the cloud and the .json must have a link to that image. However, when we tried that, the image failed to load.
2. When attempting to map Labelme's annotation format into Label Studio's annotation format, we realize that IDs are needed in Label Studio's annotation format. Meanwhile, our data is from Labelme. How can we circumvent the ID issue? Can we create our own ID schema?

Labelme's .json format can be seen here:
https://aksara-test.s3.amazonaws.com/Annotated/Bali_1.json

An example image can be seen here:
https://aksara-test.s3.amazonaws.com/Raw/Bali_1.png