perf-project / PeRF

[TPAMI 2024] PERF: Panoramic Neural Radiance Field from a Single Panorama
https://perf-project.github.io/
Other
206 stars 24 forks source link

Predicted normals are all NaN #7

Open eric11220 opened 1 month ago

eric11220 commented 1 month ago

As mentioned in #5, The output of "layer_4" is NaN for Line 71 in modules/geo_predictors/omnidata/modules/midas/dpt_depth.py _layer_1, layer_2, layer_3, layer_4 = forwardvit(self.pretrained, x)

The pre-trained checkpoint seems to load successfully for the normal prediction model. How to fix the issue? thanks!

gard-n commented 1 month ago

I describe a workaround here. It would be great if there was a solution where you didn't have to manipulate the backbone.