Difference in quantitative evaluation between your paper and LangSplat

[frankielp]

Thank you for your outstanding work!

I have a question regarding the evaluation of LangSplat in your paper on the LERF dataset compared to its original paper. The LangSplat paper reports a performance of 51.4 mIoU, whereas in your work, this figure is reported as 9.66 mIoU.

Could you kindly clarify the key differences in the evaluation methodology, dataset configuration, or other factors that might account for this discrepancy?

Thank you for your time and valuable insights!

Best regards,
Frankie.

LangSplat's Quantitative Evaluation

OpenGaussian's Quantitative Evaluation

[yanmin-wu]

Thank you for your attention.
Please refer to issue #9.