How to run on custom datasets?

[GCChen97]

Hi, I am trying to run NMF on my own datasets. My dataset format is dtu or neus, for which I write a new dataloader. But I notice the rendered results except the images of the last index are just blank with vis_every=500. I tried to tune the near_far but did not work.

Are any advice for custom datasets? For example, the scale of the scene? the size of the bounding box of the poses?

[gkouros]

I have a similar issue with my real segmented datasets. I first convert the datasets from LLFF to blender format scripts/llff2nerf.py and then train them as with other blender datasets but the metrics and reconstruction are really bad. I still haven't figured out what goes wrong.

[half-potato]

Try tuning the starting density. Is the PSNR going up?

[GCChen97]

Try tuning the starting density. Is the PSNR going up?

Hi, thanks for your advice. Which yaml file should I use for tuning the starting density? Multiple yaml files have density setting.

[half-potato]

This one. The one shown in the readme. https://github.com/half-potato/nmf/blob/main/configs/model/microfacet_tensorf2.yaml