depth = (poses[id_im-1,:3,2].T @ (point3D - poses[id_im-1,:3,3])) * sc ,WHY?

[jingfengguo-ai]

Hello, I would like to ask why the depth is calculated this way?
and What do 'sc' and 'bds_raw' mean?
What does this judgment condition mean?
if depth < bds_raw[id_im-1,0] * sc or depth > bds_raw[id_im-1,1] * sc:

I'm hoping someone can help me.

[dunbar12138]

sc is the downsampling scale of the images and bds_raw is the boundaries of the scene. The condition means we're filtering out those outlier depths which are outside the scene bounding box.

[jingfengguo-ai]

sc is the downsampling scale of the images and bds_raw is the boundaries of the scene. The condition means we're filtering out those outlier depths which are outside the scene bounding box.

Thank you for your reply! I have one more question，after running img2poses on my own dataset to get poses_bounds.npy, when I run run_nerf.py, I get negative numbers in bds_raw, resulting in sc being negative, far being negative, and near being positive. Can you give me some suggestions?
My data set was taken by drones and satellites, far from the ground.

[this233]

sc is the downsampling scale of the images and bds_raw is the boundaries of the scene. The condition means we're filtering out those outlier depths which are outside the scene bounding box.

Thank you for your reply! I have one more question，after running img2poses on my own dataset to get poses_bounds.npy, when I run run_nerf.py, I get negative numbers in bds_raw, resulting in sc being negative, far being negative, and near being positive. Can you give me some suggestions? My data set was taken by drones and satellites, far from the ground.

maybe you need modify line 66 in llff/poses/pose_utils.py：
cams[ind-1] = 1
should be modified as
cams[perm[ind-1]] = 1