Image-to-3D
Hunyuan3D-2
Diffusers
Safetensors
English
Chinese
text-to-3d

Results for Hunyuan3D on living organisms (blind human votes, with reference photos)

#63
by musharna - opened

I run Taxon3D, a blind pairwise arena where people compare two anonymised 3D models of a named
organism, with CC-licensed photos of the real thing beside them. Three Hunyuan3D checkpoints are
on the image-to-3D board. Hunyuan3D v3 scores 1044, 95% interval 1001 to 1089, over 112
head-to-head games, which puts it in the second tier of 12 models. v2 and 3.1 are also ranked.
Every head-to-head record is here:

https://taxon3d.org/models/fal:hunyuan3d-v3?c=hf-hunyuan

Every task ships the input photo, so you can see exactly what the model was given. The set is
plants, fungi and animals, which are harder than props: thin surfaces, branching, self-occlusion.

Two things that might be useful here: the failure modes are visible per task, so you can see
which organisms each checkpoint struggles with, and having three versions on the same tasks makes
the version-to-version change measurable rather than anecdotal. Code is MIT and the votes are
published on Hugging Face.

Happy to answer questions about the protocol.

Sign up or log in to comment