Ask to add individual datasets from BOP-H3 to the Hugging Face hub

[NielsRogge]

Hello @thodan 🤗

I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work through Hugging Face's daily papers as yours got featured: https://huggingface.co/papers/2504.02812.
The paper page lets people discuss about your paper and lets them find artifacts about it (your dataset for instance),
you can also claim the paper as yours which will show up on your public profile at HF, add Github and project page URLs.

I saw that BOP-H3 dataset is now available through the "bop-benchmark" Hugging Face organization.
It'd be great to have each individual datasets from BOP-H3 (HOT3D, HOPEv2, and HANDAL) listed on the Hugging Face hub as separate datasets with their own download stats. This improves their discoverability/visibility.
We can add tags so that people find them when filtering https://huggingface.co/datasets.

Would be awesome to make the dataset available on 🤗 , so that people can do:

from datasets import load_dataset

dataset = load_dataset("your-hf-org-or-username/your-dataset")

See here for a guide: https://huggingface.co/docs/datasets/loading.

Besides that, there's the dataset viewer which allows people to quickly explore the first few rows of the data in the browser.

Let me know if you're interested/need any help regarding this!

Kind regards,

Niels

[nv-nguyen]

Hi @NielsRogge,

Thank you for reaching out !

It'd be great to have each individual datasets from BOP-H3 (HOT3D, HOPEv2, and HANDAL) listed on the Hugging Face hub as separate datasets with their own download stats
Could you please clarify more as each of our datasets has its own repo and all the download stats are also available ?

Please find below the repo and download stats (April 24th 2025):

• bop-benchmark/hot3d: 10.215 downloads last month (Dataset viewer is also available)
• bop-benchmark/hope: 125 downloads last month
• bop-benchmark/handal: 113 downloads last month

Thank you,
Nguyen

[thodan]

Hi @NielsRogge , I have a questions about the download stats. On https://huggingface.co/datasets/bop-benchmark/hot3d I see HOT3D-Clips have been downloaded 9,761 times during the last month. If I understand it correctly, this number increases by 1 if someone is downloading files from this repo for 5 minutes. If that is the case, this stats seems not very informative as it depends on the user's internet connection. Isn't there a better way how to measure the number of downloads? Thanks.

[NielsRogge]

Hi,

Download stats are based on IP address: https://huggingface.co/docs/hub/en/datasets-download-stats

[thodan]

Thanks @NielsRogge for the link. Have you considered (1) counting +1 if a user downloads X MB of data instead of (2) +1 for 5 minutes of downloading? In the latter case, if a user has a very slow internet connection, they generate a high number of downloads. The first approach would be agnostic of the internet connection.

[NielsRogge]

I'll ping @lhoestq here

[lhoestq]

Hi ! That's a good point. Our system is pretty simple now, which I feel like is helpful for transparency. We also explored other options like separating downloads per dataset subsets and files but couldn't find a good tradeoff between simplicity/transparency and accuracy

[thodan]

Thanks for the reply @lhoestq !

I don't find the current approach very transparent as the download count depends on the internet speed, which is unknown.

What do you think about the approach proposed above, where +1 would be counted if a user downloads X MB (e.g. 100 MB) of the dataset? This approach doesn't depend on the internet speed and still accounts for partial downloads of the dataset (which is a good property as, e.g., in the case of BOP users rarely download full datasets).