Questions about the Table 2 in the paper,

[MinJunKang]

Hello,

In table 2 of the paper, depthsplat and mvsplat's performance are measured when P = 1.
In my understanding, these regression based gaussian splat models don't work with a single input due to cost volume architecture. (related to this issue)

How did you measure these numbers when P = 1?

Thanks

[hangg7]

Hi, this is good question. We did the simplest thing by just repeating the same image twice as input for both models. As you pointed out, these models are not intended for SV case, but we showed the evaluation results for the sake of completeness. Hope this answers your question.

[MinJunKang]

@hangg7 Thank you for your clear answer. This resolves my concern.
This is not related to this issue but could you answer another issue raised by me "Benchmark, GSO dataset's camera extrinsics are not available"? I hope to use your benchmark so I want to resolve any ambiguity existing in the benchmark. Let me close this issue after that. Thanks a lot!

[jensenzhoujh]

Hi @MinJunKang,

Looking into this and will get back to you later.

[MinJunKang]

@jensenzhoujh Any news?