When i use the following command
ns-render dataset --load-config ns_outputs/ori_splatfacto_do_dencify_distorted/splatfacto/2024-06-18_113728/config.yml --image-format png --split train+test --rendered-output-names rgb gt-rgb
for rendering images and comparing with GT.
And what I found was that there was a growing gap between the rendered view along the trajectory and the GT view.
The same goes for the ns-render camera-path command.
That looks quite strange, it seems that the frame names and poses are getting mixed up somehow. What happens if you only render test or train images, not both train+test.
Describe the bug
When i use the following command
ns-render dataset --load-config ns_outputs/ori_splatfacto_do_dencify_distorted/splatfacto/2024-06-18_113728/config.yml --image-format png --split train+test --rendered-output-names rgb gt-rgb
for rendering images and comparing with GT.
And what I found was that there was a growing gap between the rendered view along the trajectory and the GT view.
The same goes for the
ns-render camera-path
command.