search

EPFL-VILAB / omnidata

A Scalable Pipeline for Making Steerable Multi-Task Mid-Level Vision Datasets from 3D Scans [ICCV 2021]
Other
418 stars 53 forks source link

why not include scannet v2 models ? #45

Closed pengwangucla closed 1 year ago

pengwangucla commented 1 year ago

as stated questions. Does the model recoverred are bad ?

alexsax commented 1 year ago

Scannet v2 is a great dataset. We didn't include it in the initial release because it is similar to Taskonomy, and we focused our efforts on first adding more diversity (outdoor scenes, closeups of objects).

We'd happy accept pull requests :)