Datasets:

mllmTeam

/

MobileViews

like 7

Hi,

Thank you for your feedback and for pointing this out! It seems there might be an issue with the unpacking process that caused the number of samples to be fewer than expected. You could try unpacking the files again to see if that resolves the issue. Meanwhile, I’m checking the files on my end and will repack and upload them if needed. I’m also working on finding a better way for users to download, unpack, and use the dataset smoothly.

I’ll update the repository and README as soon as possible, and I’ll notify you once everything is solved. Apologies for the inconvenience, and thank you for your patience!

Thank you for providing this great dataset!

After downloading the dataset, I found that the number of samples is fewer than expected.

Running len(glob.glob(mobileviews_dir+'MobileViews_300000-522301/*.jpg')) returned 46330 not 522301-300000=222301 !!

I have double-checked the files: du -sh MobileViews_0-291197.tar.gz and du -sh MobileViews_300000-522301.tar.gz returned 12G and 11G respectively.

Could the authors please solve this missing-sample issue?@mllmTeam @GaoLongxi

Here are the counts I get (I re-extracted the tars as recommended):

• 0-291197: 52649 jpg, 52655 json
• 300000-522301: 60273 jpg, 60272 json
• Total for 112922 jpgs, there are only matching 27466 JSON files.

These are actually different from what @AutoGUI got, so there is definitely some tar issue. Would it be possible to upload in parquet format? Should be easy to do since its just two fields: image and json string.

We have already identified the tar issue, and the files have been compressed into ZIP format, which are currently being uploaded. Thanks for your advice; we will update the repository with the Parquet format as soon as possible.

Here are the counts I get (I re-extracted the tars as recommended):

• 0-291197: 52649 jpg, 52655 json
• 300000-522301: 60273 jpg, 60272 json
• Total for 112922 jpgs, there are only matching 27466 JSON files.

These are actually different from what @AutoGUI got, so there is definitely some tar issue. Would it be possible to upload in parquet format? Should be easy to do since its just two fields: image and json string.

Hi, @AutoGUI and @sohamg121 :

We’ve uploaded new .zip and .parquet versions of the MobileViews dataset to fix the issues with "data corruption" and "missing samples" in the previous tar files.

If you ran into any problems with the old files, please try downloading the new versions. We’ve also updated the README with detailed instructions for using the new formats.

Thanks for your patience, and feel free to reach out if you have any questions!