Training failure due to the scale difference between the offset and the canonical space gs.

[Time-Lord12th]

Hello，
Thanks for your great work. Have you ever encountered training failure due to the scale difference between the offset and the canonical space gs?
For example, the position range of the canonical space gs is [-0.2, 0.2], while the offset range output by the deformation model is [-1, 1]. So most of the gs will be deformed out of the camera's field of view, causing a large difference between the rendered image and the ground truth, making it impossible for the model to optimize properly.
Do you have any opinion on this issue?

[guanjunwu]

yes of course, you can just change the initialization of the deformation field in L221 and L223.

[Time-Lord12th]

thanks, but I wonder why L221 and L223 are the same. L223 should initialize m.bias?

[guanjunwu]

Yes, you can set different initialize level, but I think you can only modify L221.

[Time-Lord12th]

thanks for your advice！