Issues with custom Training Data

satyajitghana commented 1 year ago

Hi,

I'm trying to train PointNerf on my Custom Data, which is very similar to Nerf Synthetic COLMAP data, but I'm having trouble loading my point cloud

torch.Size([2989992]) torch.Size([2989992]) torch.Size([2989992])
load points_xyz torch.Size([0, 3])
Traceback (most recent call last):
  File "/mnt/thor_hdd_01/satyajit/experiments/pointnerf/run/train_ft.py", line 1081, in <module>
    main()
  File "/mnt/thor_hdd_01/satyajit/experiments/pointnerf/run/train_ft.py", line 690, in main
    _, sparse_grid_idx, sampled_pnt_idx = mvs_utils.construct_vox_points_closest(points_xyz.cuda() if len(points_xyz) < 80000000 else points_xyz[::(len(points_xyz) // 80000000 + 1), ...].cuda(), vox_res)
  File "/mnt/thor_hdd_01/satyajit/experiments/pointnerf/run/../models/mvs/mvs_utils.py", line 541, in construct_vox_points_closest
    xyz_min, xyz_max = torch.min(xyz, dim=-2)[0], torch.max(xyz, dim=-2)[0]
IndexError: min(): Expected reduction dim 0 to have non-zero size.
end loading

although the fused.ply seems fine

   ply
   format binary_little_endian 1.0
   element vertex 488944
   property float x
   property float y
   property float z
   property float nx
   property float ny
   property float nz
   property uchar red
   property uchar green
   property uchar blue
   end_header
   ]V?H,^D?@VR??g?>{K?>

I'm not sure what could be going wrong

jmchensc commented 1 year ago

Hi, I have the same problems with you when runing my own datasets which is like "load points_xyz torch.Size([0, 3])", may I ask you have found the solution to this problem?

kaichen-z commented 1 year ago

I think that this problem mainly comes from the previous filtering process consisting of several steps, you probably need to check each step for number decreasing.

iszhihao commented 5 months ago

嗨，我在运行自己的数据集时遇到了同样的问题，就像“加载points_xyz火炬。Size（[0， 3]）“，请问您已经找到了这个问题的解决方案？

请问您已经找到了这个问题的解决方案？我在更换了MVSNet的模型后，也遇到了这个问题

jmchensc commented 4 months ago

你好，后续工作没有在pointnerf基础上进行展开了

------------------ Original ------------------ From: iszhihao @.> Date: Thu,May 16,2024 11:23 AM To: Xharlie/pointnerf @.> Cc: jmchensc @.>, Comment @.> Subject: Re: [Xharlie/pointnerf] Issues with custom Training Data (Issue #78)

嗨，我在运行自己的数据集时遇到了同样的问题，就像“加载points_xyz火炬。Size（[0， 3]）“，请问您已经找到了这个问题的解决方案？

请问您已经找到了这个问题的解决方案？我在更换了MVSNet的模型后，也遇到了这个问题

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

1FUSANGSSHU1 commented 2 months ago

嗨，我在运行自己的数据集时遇到了同样的问题，就像“加载points_xyz火炬。Size（[0， 3]）“，请问您已经找到了这个问题的解决方案？

请问您已经找到了这个问题的解决方案？我在更换了MVSNet的模型后，也遇到了这个问题

您好，可以问一下个人pointnerf数据集是怎么制作的吗

Xharlie / pointnerf

Issues with custom Training Data #78