search

barbararoessle / dense_depth_priors_nerf

Dense Depth Priors for Neural Radiance Fields from Sparse Input Views
MIT License
378 stars 49 forks source link

Confused about the coordinate system and COLMAP sparse reconstruction #21

Closed BianFeiHu closed 1 year ago

BianFeiHu commented 1 year ago

Hi, thanks for your work. I want to use Scannet ground truth Camera pose to compute the depth map. However, I was confused by the coordinate system and Camera model.

  1. I found that SCANNET provides Camera to World Matrix as camera pose. Do i need to transfer it to W2C so it can be used by COLMAP?
  2. Also I don't see any code that transfer COLMAP coordinate to OPENGL coordinate, as original NeRF dose. Do you take COLMAP coordinate as model input?
  3. The camera.txt in sparse dictionary have a WIDTH, HEIGHT of 624468. However, the SCANNET original image size is 1296 968. Does this mean that we need to run COLMAP on resized images? I thought the resized work is done by extract_scannet_scene.cpp but we need COLMAP Sparse reconstruction result to run extract_scannet_scene. That is confusing. Could you help me solve these problems? Thanks a lot, Sorry I am new to NeRF and COLMAP.