Why does the nerf rebuild not work well with the blender data I created myself?

[ATing0203]

Hello, everyone. I use blender to create a set of reed blender data as shown in image01 and image 02. The command to use is:

ns-train instant-ngp-bounded --pipeline.model.background-color white --experiment-name luwei blender-data --data data/blender/luwei --scale-factor 0.23 .

The result is shown in image03. But there is no color. And the exported point cloud is shown in image04 and image05, with a circle of something like a camera path. Does anyone know why that is? What should be done to fix it? I would be very grateful for your answer!

[amirbarda]

Did you figure it out? I am getting the same problem using nerfacto , the reasons seems to be --pipeline.model.background-color white , but when I remove it, it fails to render anything except a white background. weirdly, when trying "lego" in the synthetic blender data with --pipeline.model.background-color white , it works perfectly.

[ATing0203]

Did you figure it out? I am getting the same problem using nerfacto , the reasons seems to be --pipeline.model.background-color white , but when I remove it, it fails to render anything except a white background. weirdly, when trying "lego" in the synthetic blender data with --pipeline.model.background-color white , it works perfectly.

Not solved, but it feels like a scene frame problem. Maybe the center of the scene we modeled isn't the same as lego.