search

GhiXu / Geo-Neus

Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction (NeurIPS 2022)
MIT License
290 stars 18 forks source link

performance in DTU #24

Open JuliusQv opened 8 months ago

JuliusQv commented 8 months ago

Thanks for your excellent work.

I used the default parameter settings in the repo，and got a very different performance on the DTU dataset worse than in the paper.

scan24 0.446 , but your paper is 0.375. scan37 0.968，but your paper is 0.537.

Do you have any suggestions？

JuliusQv commented 7 months ago

Thanks for your excellent work. I used the default parameter settings in the repo，and got a very different performance on the DTU dataset worse than in the paper. scan24 0.446 , but your paper is 0.375. scan37 0.968，but your paper is 0.537. Do you have any suggestions？

Hello, I tried to evaluate the trained DTU model and ran eval. py, but received an error (ValueError: Found array with 0 sample (s) (shape=(0,3)) while a minimum of 1 is required by NearestNeighbors.). May I ask how you obtained this result with scan24=0.446 and what code was used?

I use the default repo. Maybe you can check if the predicted mesh or gt mesh is empty.