Size discrepancy for PlantInquiryVQA dataset
Hi,
After fully downloading SyedNazmusSakib/PlantInquiryVQA, the total size is ~1.5GB locally, whereas the document mentions the image corpus is about 3.5GB.
All visible files on the repository have been downloaded without interruption.
Could you clarify whether part of the image data is missing on HF hub, or the 3.5GB refers to uncompressed original size?
Thanks a lot.
Hello,
Thanks for pointing this out.
Could you please check whether the total number of downloaded images/files matches the number available in the Hugging Face repository? If the image and file counts match, then the difference may be related to file compression, storage format, or how the dataset size was originally calculated. However, if the number of images/files is lower than expected, there may have been an issue during the original upload.
Please let us know what file/image count you see on your side. If there is any discrepancy or missing data, we would be happy to verify the repository and provide the complete dataset again if necessary.
Thanks!
Hello,
Thanks for pointing this out.
Could you please check whether the total number of downloaded images/files matches the number available in the Hugging Face repository? If the image and file counts match, then the difference may be related to file compression, storage format, or how the dataset size was originally calculated. However, if the number of images/files is lower than expected, there may have been an issue during the original upload.
Please let us know what file/image count you see on your side. If there is any discrepancy or missing data, we would be happy to verify the repository and provide the complete dataset again if necessary.
Thanks!
Hello,
I checked the Hugging Face repository by enumerating all files under the images/ directory via the repository API.
The repository currently contains exactly 10,000 image files, with a total size of 1,474,667,156 bytes (approximately 1.37 GiB / 1.47 GB).
My local download also contains exactly 10,000 images and has the identical total byte size, so the download appears complete and the discrepancy is not caused by compression or local storage calculation.
However, the paper states 24,950 images and approximately 3.5 GB. In addition, the released train/test CSV files reference 24,964 unique image IDs, of which 14,964 do not have a corresponding image in the current repository.
Therefore, it seems that the Hugging Face image upload is incomplete. I would appreciate it if you could verify the repository and re-upload the missing image files.
Best regards,