search

omniobject3d / OmniObject3D

[ CVPR 2023 Award Candidate ] OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation
462 stars 12 forks source link

Camera conventions #25

Closed akanimax closed 1 year ago

akanimax commented 1 year ago

Hello,

could you please provide information on the camera conventions used with this dataset? I am trying to use this dataset in PyTorch3D, but not getting correct 3D reconstructions even though the cameras look correct in cam-visualisations.

Best regards, @akanimax

louhz commented 1 year ago

Same Question

wutong16 commented 1 year ago

Hi. We are using the same format of camera parameters as the NeRF_synthetic dataset, following the OpenGL coordinate system (x: right, y: upward, z: backward), which indeed differs from Pytorch3D's coordinate system. Please find their differences and conventions here.

akanimax commented 1 year ago

Hi @wutong16, thanks for the comment. This is super helpful 👍.