Conversation
Qwlouse
left a comment
Qwlouse
left a comment
There was a problem hiding this comment.
Hey, this looks pretty cool, thank you so much for putting this together, and apologies for taking so long to review. I've added a few minor comments. Once they are addressed I'd be happy to merge this.
|
|
||
|
|
||
| def write_coordinates_batch(data, directory, file_template="object_coordinates_{:05d}.png", | ||
| def write_coordinates_batch(data, directory, file_template="object_coordinates_{:05d}.tiff", |
There was a problem hiding this comment.
please do not change from png to tiff here. That is a breaking change.
If you need coordinates as a tiff file, then add a new write_coordinates_batch_tif method and in your worker set kb.file_io.DEFAULT_WRITERS["object_coordinates"] = write_coordinates_batch_tif
| missing_textures = sorted({img.filepath for img in bpy.data.images | ||
| if tuple(img.size) == (0, 0) or len(img.filepath)>1}) | ||
| if missing_textures and not ignore_missing_textures: | ||
| raise RuntimeError(f"Missing textures: {missing_textures}") |
There was a problem hiding this comment.
Any particular reason for inlining _check_missing_textures here?
| cloudml-hypertune | ||
| google-cloud-storage | ||
| imageio | ||
| imageio-ffmpeg |
There was a problem hiding this comment.
I don't think this should be a mandatory kubric dependency just because the dynamic nerf worker uses it. I'd prefer adding specific instructions for additional dependencies to that worker.
2 participants
add dynamic NeRF challenge