The experimental results are inconsistent with the results in the paper.

[Yuhuoo]

I have run the command as described in your project without modifying any code, but the results obtained are different from those in your paper.

python train.py -s {dataset_path} --exp_name {exp_name} --eval --ours	scene	PSNR	SSIM	LPIPS
train	20.96	0.76	0.27
truck	22.92	0.82	0.21
drjohnson	28.94	0.89	0.27
playroom	30.09	0.90	0.26

python train.py -s {dataset_path} --exp_name {exp_name} --eval --DSV	scene	PSNR	SSIM	LPIPS
train	20.68	0.77	0.26
truck	20.08	0.75	0.25
drjohnson	28.74	0.89	0.26
playroom	28.53	0.89	0.27

The result in your paper:

[crepejung00]

Hi, We tested our codes multiple times and verified that the codes faithfully reproduce the results on our paper. Although we haven't changed any random seeds for our experiment, due to the randomness in the PDF sampling in the split process of the adaptive density control of 3DGS, the results can differ about +- 0.2dB PSNR from the reported values. It seems that the results of all other scenes except the truck scene in the Tanks & Temples dataset are reasonable. In addition, we have re-cloned our repository and tried training the scenes you mentioned, and we have found that all scenes match the results on the table. Here are the results:	scene	PSNR	SSIM	LPIPS
train	21.089	0.756	0.271
truck	23.478	0.833	0.200
drjohnson	28.960	0.893	0.268
playroom	30.139	0.902	0.258

Could you please try training the truck scene one more time and check if the dataset has not been modified from the original version, which is provided by Inria 3DGS repo?

Thanks, Jaewoo.