search

lukasHoel / text2room

Text2Room generates textured 3D meshes from a given text prompt using 2D text-to-image models (ICCV2023).
https://lukashoel.github.io/text-to-room/
MIT License
1k stars 69 forks source link

Point Cloud Representation #10

Closed jaidevshriram closed 1 year ago

jaidevshriram commented 1 year ago

Hey, love this work! I was wondering why you preferred a mesh representation over point clouds - why not just create the mesh after the entire generation process is done?

Caoang327 commented 1 year ago

Thanks for your interest! We tried point cloud representation initially and found it led to more severe distortions and ill-structured inpainting masks.

Point Cloud is a more discrete representation compared to mesh, and even rendering a flat region/plane represented by a point cloud from an extreme viewpoint may lead to many missing regions inside this plane, while mesh could solve this problem appropriately