Looks all good !

Who are the source language producers?

This section is N/A for vision datasets - this is a NLP thing. You can remove it.

Also, not super sure about the schema to put under Data Instances since the input image and the depth map are primarily generated from an h5 file.

You did right by showing what an example dictionary looks like

Thanks!

Two final things:

• https://huggingface.co/datasets/sayakpaul/nyu_depth_v2/discussions/6 (probably can be tackled later?)
• First question raised in https://github.com/huggingface/datasets/issues/5255#issuecomment-1339013527

Yea I think it's fine to focus on the two tasks we support. You can also mention the other tasks but without detailing them if you want

Done adding those.

Merging away.