Download SpaceType-Wise Data

[ashwin-ned]

Hello and thank you for this fantastic and detailed work. It addresses key issues with depth estimation datasets in Sota research.

Are there any download scripts available that let users download only specific parts of the dataset according to the SpaceTypes mentioned in the paper?
For example,e I would like to fine-tune DepthAnything only on depth ranges from 5-15m and would only want to download the Lounge, Hallway, and Large Room Space Types.

[DepthComputation]

Hi, Thank you for the interest in this work. Sorry Currently there is no script available for partial data download. One may need to download the whole dataset and fetch specific category by tags. Ashwin Nedungadi ***@***.***>於 2025年1月15日 週三，下午7:59寫道：

…

Hello and thank you for this fantastic and detailed work. It addresses key issues with depth estimation datasets in Sota research. Are there any download scripts available that let users download only specific parts of the dataset according to the SpaceTypes mentioned in the paper? For example,e I would like to fine-tune DepthAnything only on depth ranges from 5-15m and would only want to download the Lounge, Hallway, and Large Room Space Types. — Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BAJTBJUXUS7FLUT5J6RZJ3T2KZEQXAVCNFSM6AAAAABVHCDJPSVHI2DSMVQWIX3LMV43ASLTON2WKOZSG44DSNRRHE4DOMI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[ashwin-ned]

Thank you for the prompt response. Would it be possible to know which among the 8 data chunks contains the large hallway, lounges, and library space types? If they are all randomized and not ordered in any particular way I will go ahead and close this issue.
Thank you again.

[choyingw]

Hi,

The data is split by diving the whole zipped dataset, and each chunk cannot be unzipped without the whole 8 zip files. The only way is to download all the zip files and fetch the category you want by metadata