codes doesn't run successfully when computed normals

[hey0chen]

loaded output/living_room/full_trajectory/a_living_room_with_a_lit_furnace,couch/no_input_image_file/2023-05-22_20:05:49.520666Z/fused_mesh/after_generation.ply computed normals

I think this problem maybe caused by pymeshlab

[lukasHoel]

What's the error that you get? You can try to run this step in the GUI of Meshlab manually.

[hey0chen]

What's the error that you get? You can try to run this step in the GUI of Meshlab manually.

The code doesn't seem to run anymore

and I get the following error when I use meshlab manually

[lukasHoel]

• You need to compute the normals manually in Meshlab as well before running the Poisson step

• For me, poisson surface reconstruction usually finishes in ~10 minutes. Maybe #5 is also related, could you check if it is a similar issue by trying with the GUI on your local machine?

[hey0chen]

Thanks for your answer, I successfully manually running Poisson step with the GUI on mt local mechine, it seems to be the machine's problem, I will check it

[zf-666]

I meet the same problem, have u solve it? Thank u

[hey0chen]

There seems to be a problem in this part of the code ms.generate_surface_reconstruction_screened_poisson(depth=depth) The parameter "depth" affects the speed,I changed depth=6 and tried many times, only one worked but it was still very slow.

[zf-666]

I found this step to run on my A100 server for 34163s, nearly 10 hours, but yes it does work.

[lukasHoel]

I found this step to run on my A100 server for 34163s, nearly 10 hours, but yes it does work.

That is definitely way too long for meshes with about 3M vertices. Double check if all CPUs are used and maybe try different versions of pymeshlab or an alternative implementation (e.g. Open3D).