henry123-boy / level-s2fm_official Goto Github PK

Hello,

I searched vren module but I can't figure it out.
How to install vren module?
Thank you

Hi, please could you include a license?

Hi,

Thanks for the amazing work. I wonder if you could provide the reconsturction results for all the scenes in ETH3D datasets?

hello! could you tell me where I could find camera.npz in the data :ETH3D?

Hi,
Thank you for sharing your awesome work!

Could you please share how much time it took to train the full network and the hardware you used?
I couldn't find any additional explanation about the training time or hardware.

Thanks!

Hi, thanks for your work and contribution to SFM community!

I have 2 questions about cameras.npz:

cameras.npz is the file which is referenced at

1. In your provided DTU dataset, only scan65 has cameras.npz while other scenes(like scan24) don't provide, so if we want to train on other DTU scenes(like scan24) , the program will run into FileNotFoundError: [Errno 2] No such file or directory: 'levels2fm-opendata/DTU/scan24/cameras.npz'

2. what's the meaning of scale_mat & world_mat? From the code, I can tell that projection matrix is the multiplication of scale_mat and world_mat, and the projection matrix is decomposed into intrinsic and extrinsic(pose) later. The world_mat itself is also decomposed into intrinsic and extrinsic(pose_org) , which is written into txt file. I am not clear with the difference between pose and pose_org

Can you provide me with some help? Thanks again!

Hi, thanks for your excellent work.

“During our evaluation, we used the provided API of Reconstruction Align in COLMAP [34] to do that.” I see this in your paper, but I can't find relevant API of colmap in its doc.

Can you tell me the command that you use with colmap ?

Thanks!

Hi,
Very interesting work, thanks for open-sourcing the code. I downloaded the data and I run the following command:

python train.py --group=DTU --pipeline=LevelS2fM --yaml=DTU --name=65_dual --data.dataset=DTU --data.scene=scan65   --sfm_mode=full  --Ablate_config.dual_field=true

The training generates 00000045.ply, and I saw the following error:

Traceback (most recent call last):                                                                                                              
  File "/media/HDD_2TB/amughrabi/projects/Level-S2FM_official/train.py", line 23, in <module>
    main()
  File "/media/HDD_2TB/amughrabi/projects/Level-S2FM_official/train.py", line 20, in main
    m.train(opt)
  File "/media/HDD_2TB/amughrabi/projects/Level-S2FM_official/pipelines/LevelS2fM.py", line 245, in train
    new_id = pose_graph_left[0]
IndexError: list index out of range

Would you please advise?

DTU dataset provides groundtruth point clouds, LevelS2FM also generates mesh. I wonder the way that you evaluate mesh quality.

From my point of view, the process is compromised by 3 steps:

• first step is to sample points from generated mesh;
• secondly, aligning sampled point clouds with DTU ground-truth point clouds (this alignment including 3 parameters, rotation, translation, and scale) . I guess this step is implemented with open3D.registration ?
• lastly, compute chamfer distance between point clouds

Is my understanding correct? Can you provide corresponding codes of evaluating mesh quality ?

Hi, thanks for the great work!
I tried to use your method to get a better camera pose on my own dataset, but failed. It raised a nan bug in sphere_tracing. After investigation, I found the method embed_fn and SDF_MLP will output nan.

My dataset only contains 30 images and can be reconstructed well on NeRF. I suppose the camera pose from colormap is generally correct. I follow preparation/README.md to create my dataset, and save intrinsics.txt and pose(Is it necessary? or just identity matrix?) from colmap output.

I wonder where might be wrong and how can I create my own dataset.
The other question is what's the meaning of the output/mesh/cam000xxx and how to output the camera pose or convert the optimized camera pose to colmap style.

Hi, excellent work, we all love it!

I see this page provides instructions for creating custom dataset: eabdd11

but it's not easy to understand purpose of two_view.npy & N_views.npy only from codes without documents, so hope you can share us more descriptions about the 2 files!

THANKS FOR YOUR INCREDIBLE WORK!