@DeepGHS
Account based inNetherlands!
About this account
- Account based in
- Netherlands
- Connected via
- United States App Store
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
DeepGHS is a passion-driven, non-profit community building anime/2D-focused infrastructure. Contact us [email protected]
Earth
Joined October 2025
- Tweets6
- Following12
- Followers110
- Likes13
Pinned Tweet
π¨ DeepGHS β Building the future of anime-tech, one dataset at a time!
π 100% non-profit | π open-source | π§ powered by anime lovers
Founder @narugo1992
Discord Community discord.gg/EAW4WqFdKY
Explore β github.com/deepghs | huggingface.co/deepghs
This project was led and trained entirely by @IrisRainbowNeko and Lamb Bao
, with assistance from @narugo1992
DeepGHS retweeted
It looks like HF is closing the download access to private repo of who(user/org) have exceed their private storage quota
So, it means even if you want to move the data to other place for free up space you still need to pay at first.
πBig update to our Animetimm project on huggingface! danbooru tagger adds ConvNeXtV2 Huge & EVA02 Large models. ConvNeXtV2 Huge now achieves leading performance (~0.58 macro F1, ~0.697 micro F1) .
With the new tag-level threshold feature, you can now get more accurate results without tuning the threshold!
More new taggers: e621tagger, Pexel tagger & more to meet all your tagging needs!
. Explore 20+ models & metrics: huggingface.co/spaces/animetβ¦
. More improvements coming!
AnimeTimm (PyTorch Image Models For Anime) is a project started by DeepGHS, @narugo1992 and SmilingWolf to build foundational infrastructure for anime image processing.
Our vision goes beyond just tagging - we aim to provide pre-trained models that can serve as the foundation for various anime-related tasks. As new excellent network architectures emerge, we will continue to add more anime pre-trained models to our collection. These models are designed to be versatile building blocks that the community can leverage for their own applications, not just limited to tagging purposes.