search

KAIST-VICLab / cloud_nerf

MIT License
50 stars 19 forks source link

Did you try experimenting with a different loss function than the one used in this experiment? #20

Closed olzlof51 closed 10 months ago

olzlof51 commented 10 months ago

I'm curious about which functions were used and how the results turned out.

KAIST-VICLab commented 10 months ago

We tried L1 and L2 for the color supervision. And the typical L2 loss gives better rendering results.