search

changwoonchoi / 3Doodle

Official implementation of 3Doodle: Compact Abstraction of Objects with 3D Strokes (SIGGRAPH 24', Journal track)
https://changwoonchoi.github.io/3Doodle/
Other
71 stars 6 forks source link

How to render own object to create dataset? #6

Open yiboz2001 opened 1 month ago

yiboz2001 commented 1 month ago

Hi! How to render own object to create dataset? Do you have any script?

changwoonchoi commented 1 month ago

Hi, are you asking how to render images of objects to feed 3Doodle? If so, we used blender for rendering, and sampled random poses on upper hemisphere of the object. The script to generate images are almost similar to the original NeRF's blender python script.

yiboz2001 commented 1 month ago

Thanks for reply! And could you provide any script or guidance about creating the SfM / Line SfM points? It looks like very important.

changwoonchoi commented 1 month ago

Since we borrowed SfM-related code from other colleague, we cannot offer the code currently. But basically, we used COLMAP to triangulate feature points to obtain sparse point clouds with known camera poses for synthetic dataset. And for Line SfM in our ablation study, we used EMAP.

DanielCho-HK commented 2 weeks ago

Since we borrowed SfM-related code from other colleague, we cannot offer the code currently. But basically, we used COLMAP to triangulate feature points to obtain sparse point clouds with known camera poses for synthetic dataset. And for Line SfM in our ablation study, we used EMAP.

Dear author, I am trying to use my own data to test this model and would like to ask how to get the camera pose information saved in a JSON file like the one in your provided data. I have used the Blender tool to render multi-view images of a 3D model and would like to use ColMap to compute the camera pose of each image. But I need help finding a way to do so. Thank you for your suggestions.