Thank you for the amazing contributions made by your team. I am also very interested in using Neuralangelo for indoor reconstruction. I have successfully used Neuralangelo to reproduce a meeting scene, but there are still some differences compared to the results in the paper. I have the following questions:
1.The results of local reproduction are not satisfactory in terms of reconstructing texture-lacking floors, as there are many holes present. In contrast, the floors in the paper are very smooth. Do you have any suggestions?
The performance of self-built datasets is not ideal, and there is a lot of noise in the reconstruction, especially in complex indoor environments.The self-built dataset was collected in the laboratory and contains many chairs and computers. It was trained on Neuralangelo and InstantN-NGP, and the synthesis of new viewpoints appears to be normal. However, the reconstruction results are not satisfactory.
I followed the data preprocessing methods described in the documentation and used Blender to crop the bounding boxes to the desired reconstruction scene. Both of them rendered well, with a PSNR of 27. However, the reconstruction result of the Meetingroom scene is noticeably better than that of the self-built dataset. I would like to know the possible reasons for this difference.
I would be extremely grateful if I could receive your valuable suggestions.
Thank you for the amazing contributions made by your team. I am also very interested in using Neuralangelo for indoor reconstruction. I have successfully used Neuralangelo to reproduce a meeting scene, but there are still some differences compared to the results in the paper. I have the following questions:
1.The results of local reproduction are not satisfactory in terms of reconstructing texture-lacking floors, as there are many holes present. In contrast, the floors in the paper are very smooth. Do you have any suggestions?![Screenshot from 2023-10-10 16-04-51](https://github.com/NVlabs/neuralangelo/assets/49439134/60c658a2-2821-4394-ab65-2867ed0da8f5)
I would be extremely grateful if I could receive your valuable suggestions.