Hi,
First of all, thanks for sharing such a great work.
We've trained a mip-nerf 360 model with our own images.
And rendered images with the pose of original training data (We've copied dataset.poses to dataset.camtoworlds)
We expected the rendered images exactly aligned with training images, but we've got this as shown below which show slight misalignment.
Could you tell us which part we've done wrong, and how to fix it?
Hi, First of all, thanks for sharing such a great work.
We've trained a mip-nerf 360 model with our own images. And rendered images with the pose of original training data (We've copied dataset.poses to dataset.camtoworlds) We expected the rendered images exactly aligned with training images, but we've got this as shown below which show slight misalignment. Could you tell us which part we've done wrong, and how to fix it?