-
Notifications
You must be signed in to change notification settings - Fork 52
Description
Thank you for your great work!
I am currently reproducing the results from the NeuS2 paper using the code you provided, but am running into issues with 3D human reconstruction. Specifically, when using a custom human dataset, the reconstructed model has a shell enveloping the body.
I used colmap to get the corresponding transforms.json file and started training using the following command python scripts/run.py --scene data/Thuman2.0/124/transforms.json --name bmy-124-base --network base.json --n_steps 15000. The final generated mesh is as follows.

I also tried to use dtu.json to train, but the results were not much different.
Could you provide some guidance or insight into why this might be happening? Any suggestions on how to resolve this or further debugging would be greatly appreciated.