Image-to-3D
Trellis
English

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

#27
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. TRELLIS 2 is currently in the
top tier of the image-to-3D board: Bradley-Terry score 1067, 95% interval 995 to 1137, over 51
head-to-head games. Its record against every opponent it has met is here:

https://taxon3d.org/models/replicate:trellis2?c=hf-trellis

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 it struggles with, and if a newer checkpoint exists I will run it on the same
tasks. 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