Question about table.1 in the paper

[hebing-sjtu]

I am intrigued by the research presented in your paper. While going through the document, I observed that PSNR corresponding to mesh is provided in Table.1, but there seems to be a missing entry for Gaussian. From my understanding, mesh and Gaussian have a one-to-one correspondence and can be independently rendered to generate images. I was wondering if the authors could kindly explain why Gaussian's PSNR data is not included in the table. This information would be valuable for evaluating the method's performance and making comparisons. Any clarification on this matter would be greatly appreciated.

[Net-Maker]

Same question

[Isabella98Liu]

Hi @hebing-sjtu and @Net-Maker, thank you for your question! Yes, we do have Gaussian rendering images. Due to the anchoring, our Gaussians are more evenly distributed across the surface and tend to have more uniform scales. Consequently, our method uses fewer Gaussian points than the standard 4DGS, which might generate many small Gaussians to enhance rendering quality in difficult areas. Additionally, since our focus is on mesh reconstruction, we do not explicitly compare our Gaussian rendering quality with other methods. :) Let me know if you have any other thoughts!

[tb2-sy]

Hi @hebing-sjtu and @Net-Maker, thank you for your question! Yes, we do have Gaussian rendering images. Due to the anchoring, our Gaussians are more evenly distributed across the surface and tend to have more uniform scales. Consequently, our method uses fewer Gaussian points than the standard 4DGS, which might generate many small Gaussians to enhance rendering quality in difficult areas. Additionally, since our focus is on mesh reconstruction, we do not explicitly compare our Gaussian rendering quality with other methods. :) Let me know if you have any other thoughts!

Hi, thanks for your nice work! I run your code on the Jumpingjacks scene of d-nerf, and the PSNR of the Gaussian image on the testset was 28.6, and the Mesh image PSNR was 31.91. Is this reasonable?

[Isabella98Liu]

Hi @tb2-sy, yes, this should be normal.