search

hzxie / CityDreamer

The official implementation of "CityDreamer: Compositional Generative Model of Unbounded 3D Cities". (Xie et al., CVPR 2024)
https://haozhexie.com/project/city-dreamer
Other
589 stars 39 forks source link

The output is video, can export the building structure,mesh obj or other formats... #17

Closed kuaiqushangzixiba closed 3 months ago

kuaiqushangzixiba commented 3 months ago

The output is video, can export the building structure,mesh obj or other formats...

hzxie commented 3 months ago

You can use COLMAP to reconstruct 3D meshes from the generated video.

kuaiqushangzixiba commented 3 months ago

Ok, so citydreamer is actually a photo, but from all angles, can I understand it this way, is it similar to 3dgs?

hzxie commented 3 months ago

Yes. 3D-GS is an explicit representation, while CityDreamer uses an implicit representation NeRF.

kuaiqushangzixiba commented 3 months ago

How do the video sequence consists of 600 frames with a resolution of 960×540, I took the osm data provided and generated a video of only 6 seconds

hzxie commented 3 months ago

Change this value to 600

https://github.com/hzxie/CityDreamer/blob/49a0d9e05908861fdda2cf4d73ef32ac4309d553/scripts/inference.py#L58