Just obtained mesh of a part of object

[HungNgoCT]

Hi,

I trained neus-facto, then extract mesh of a person. However, just got the mesh of the belly as video below. I understand that I should scale the scene. Can you tell me which parameter to scale the scene to make it smaller?

https://github.com/autonomousvision/sdfstudio/assets/110378166/14dd629f-94c5-479f-8b72-32300eb4e236

During training time, the rendering is good, as shown in the picture below

Reproduce:

1. Capture a video with my phone
2. run ns-process-data
3. process_nerfstudio_to_sdfstudio.py
4. run extract_monocular_cues.py
5. ns-train neus-facto
6. ns-extract-mesh

Thank you

[dlmuzpy]

I guess that missing part of the object may come from "process_nerfstudio_to_sdfstudio" process. If you set "--include_mono_prior True", your data will be cropped and resized to 384*384, causing some image patches removed. By the way, could you share your detailed commands on "ns-train neus-facto"? I am also researching how to reconstruct a person object.

[shenggedeqiang]

Hi,

I trained neus-facto, then extract mesh of a person. However, just got the mesh of the belly as video below. I understand that I should scale the scene. Can you tell me which parameter to scale the scene to make it smaller?

bandicam.2023-09-15.15-36-54-583.mp4 During training time, the rendering is good, as shown in the picture below

Reproduce:

1. Capture a video with my phone
2. run ns-process-data
3. process_nerfstudio_to_sdfstudio.py
4. run extract_monocular_cues.py
5. ns-train neus-facto
6. ns-extract-mesh

Thank you

can you share your parameters of running these command?