NVlabs / instant-ngp

Instant neural graphics primitives: lightning fast NeRF and more
https://nvlabs.github.io/instant-ngp
Other
16.06k stars 1.94k forks source link

Poor performance using custom dataset results #1527

Open yuanjie0217 opened 8 months ago

yuanjie0217 commented 8 months ago

I am trying to use a custom data set to train on the instant-ngp project. My data set is simulated building data. This is simulated based on the SAR imaging principle. I added a mask to the simulation results to make fewer impurities. Then I estimated the camera pose through colmap and made my own data set with reference to the project's json file. However, the performance of my data set on instant-ngp was very poor. How can I improve it to improve the final result?

1710419179839

000020 ![Uploading 1710419400269.png…]()

yuanjie0217 commented 8 months ago

![Uploading 1710419400269.png…]() This is the final rendering result