I was playing around with Garfield a bit and could get it to train successfully on colmap / MiPNeRF360 data (see the screenshot from the kitchen scene at the bottom), but when I try training on the original Blender data from the NeRF paper, it doesn't seem to converge and i get the following output:
The train cmd I used:
ns-train garfield blender-data --data /path/to/lego/data
I also tried processing the blender data into colmap format with ns-process-data, but colmap fails and only matches 2 images, so that doesn't seem to work either.
Is training on Blender data not supported? Or any idea on what I'm doing wrong? Thanks :)
Hi, thanks for open sourcing this cool repo!
I was playing around with Garfield a bit and could get it to train successfully on colmap / MiPNeRF360 data (see the screenshot from the kitchen scene at the bottom), but when I try training on the original Blender data from the NeRF paper, it doesn't seem to converge and i get the following output:
The train cmd I used:
ns-train garfield blender-data --data /path/to/lego/dataI also tried processing the blender data into colmap format with ns-process-data, but colmap fails and only matches 2 images, so that doesn't seem to work either.
Is training on Blender data not supported? Or any idea on what I'm doing wrong? Thanks :)