About how to reproduce the paper.

[DavidXu-JJ]

Dear authors,

Thanks for your nice work. The project is nice. Congrats for your work!

I wish to reproduce your work, but I meet two problems here. Hoping to get your notice and reply.

1) Could you please give detailed guidance on how to preprocess our own data?

README doesn't seem to detail how to split the train and test data.

The original paper seems to say only pure test-time optimization is needed.

So, if I wish to do an evaluation, do I only need to run directly with the test data?

2) The README only gives how to inference.

But no quantitative evaluation code is given(If I haven't missed it).

If that's convenient, could you please offer guidance to reproduce the quantitative results in the paper's table?

Thank you in advance! Best, Jingwei

[DavidXu-JJ]

For question 1, I wish to additionally ask that:

As is mentioned in the paper:

While using neural methods, INR is a pure test-time optimization.

From my understanding, do we only need to make the inference on each input file?

Inference one point cloud once at a time. Am I right?

Thank you!

[DavidXu-JJ]

No reply, closed.