Release AnyDepth dataset and link models on Hugging Face

[NielsRogge]

Hi @SteveZeyuZhang 🤗

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/2601.02760.
The paper page lets people discuss your paper and find artifacts about it (your models for instance). You can also claim the paper as yours, which will show up on your public profile at HF, and add GitHub and project page URLs.

It's great to see that you've already hosted the AnyDepth model checkpoints on the Hub at https://huggingface.co/AIGeeksGroup/AnyDepth! To improve their discoverability, you can link them to the paper page by adding arxiv: 2601.02760 to the model card metadata (guide here).

I also noticed the importance of your data-centric approach and the curated 369K filtered dataset. Would you like to host that dataset on https://huggingface.co/datasets as well? Hosting on Hugging Face will give it more visibility and allow people to easily load it via the datasets library:

from datasets import load_dataset

dataset = load_dataset("AIGeeksGroup/anydepth-filtered-dataset")

If you're interested, you can find a guide for dataset uploading here. We also support a dataset viewer that allows people to explore samples directly in their browser.

Let me know if you're interested or need any guidance!

Kind regards,

Niels
ML Engineer @ HF 🤗

[chenhaomingbob]

same question

[Zephyr0609]

Hi, we have released our training dataset, and you can read these txt files for training!

[NielsRogge]

Hi @Zephyr0609, thanks for releasing the training dataset files!

I noticed that the models at https://huggingface.co/AIGeeksGroup/AnyDepth are not yet linked to the paper page. Could you link them by mentioning the paper URL (https://huggingface.co/papers/2601.02760) in the model card? This will make the models automatically show up on the paper page.

See also here for more information: https://huggingface.co/docs/hub/model-cards#linking-a-paper

Regarding the dataset, I see you have released the training split txt files in the GitHub repo. Would you also consider hosting the actual dataset (images + depth maps) on https://huggingface.co/datasets? This would give it more visibility and allow people to easily load it via the datasets library, as mentioned in the original issue.

Thanks!