will-zzy

I identified a bug: in forward.cu, lines 993, 1002, 1031, and 1035, `T` should be `corner_Ts[0]`. After correcting this in the code, the results of the two functions became consistent.

> Hi, thanks for reporting the bug. It is fixed in the latest commits. > > I haven’t test it on 2DGS but I think GOF can also be applied...

> Hi, thanks for sharing the results. Could you be more specific for what you did? Are you using the pretrained model of 2DGS to extract the mesh you show...

> Hi, defining the opacity field for 2D Gaussian is simpler than what you implemented. I mean something like this  It is a step function that is zero before...

> @will-zzy I believe that the proposed technique can be applied to 2DGS as well. There is no evidence that the GOF needs a 3D blob (nor did the paper...

@niujinshuchong @hbb1 Thank you for your responses! My questions have been resolved, so I will now close this issue.

@niujinshuchong Certainly! In my testing, using 2DGS with DMTet+ binary search consistently failed to yield satisfactory results, which I believe is related to the initial number of points. To illustrate...

You should compile tetra-triangulation(cmake . and make) and pip install it.