search

Harry-Zhi / semantic_nerf

The implementation of "In-Place Scene Labelling and Understanding with Implicit Scene Representation" [ICCV 2021].
Other
430 stars 55 forks source link

Generating the traj_w_c file for custom dataset #40

Closed AadeshVarude closed 1 year ago

AadeshVarude commented 1 year ago

Thank you for your great contribution and for sharing this wonderful code.

I am trying to run the semantic nerf on my custom dataset where I have a bunch of images on which I wish to train this network. How do I generate the traj_w_c file as well as other required files?

Thank you in advance for your help.

Harry-Zhi commented 1 year ago

Hi. We have updated the data generation codes in the repo. The traj_w_c is camera to world camera poses under OpenCV convention. The codes will transform it properly to OpenGL-camera to Habitat-World-coordinate.

As to generate this traj_w_c, a straightforward way is to randomly sample N camera positions and N look-at points (which corresponds to N discrete camera poses) before smoothly interpolating spatially to, for example, 5N or 10N poses.