kxhit / vMAP

[CVPR 2023] vMAP: Vectorised Object Mapping for Neural Field SLAM
https://kxhit.github.io/vMAP
Other
325 stars 20 forks source link

data generation for other replica scenes #18

Closed Taekbum closed 1 year ago

Taekbum commented 1 year ago

Hello. I appreciate you sharing code of this impressive research. I'm now interested in evaluating the performance of vMAP in different replica scenes, so I'd like to ask about the data generation process. First, how were the trajectories for model training (trajectory 00) and novel view synthesis (trajectory 01) generated? Second, how to obtain object-level meshes ? (to evaluate the 3D reconstruction performance at the object level) The original replica dataset seems to provides only scene-level meshes. Thank you!

kxhit commented 1 year ago

Hi, thanks for your interest in our work!

  1. The rendering data generation is two steps: a. trajectory generation, b. Habitat Simulator rendering. We use the code of Semantic-NeRF for generating the trajectories. It seems the author @Harry-Zhi hasn't released data generation code yet. For habitat rendering you may find some related code. If you have the pose file, I can help you rendering.
  2. For the object instance 3D meshes extraction, I followed demo code here .
Taekbum commented 1 year ago

Thanks for your help! But I have more questions.. Now I generated trajectory files using code similar to MoreFusion. However, I don't know how to use this code to render images given pose file. The code seems to use random pose. Can you help me? Or would it be possible for me to provide you with a pose file and request rendering? Thank you so much!

kxhit commented 1 year ago

Hi @Taekbum, you can send me your pose file via email and see if I can get it rendered correctly!

Taekbum commented 1 year ago

Thank you so much!

kxhit commented 1 year ago

Hi @Taekbum I just upload replica data generation scripts. Could you try and let me know if it works well? Cheers!

Taekbum commented 1 year ago

@kxhit It works well. Thank you!