Hello, Thanks for your nice work!
I want to know the hyper parameters settings that correspond to the results in your paper. I noticed that you set the iterations to 2000 and opacity reset interval to 200 in your released code. but I didn't get the same good results in T&T dataset as these you show in your paper, especially in outdoor scenes where the background is usually blurred without feature matching (Even in training views). I'm not sure if the opacity reset interval was set too small. Do you have some suggestions?
Here are some of my results, you can clearly find the render quality of background is bad ,which is different from the results in your papar.
In indoor scenes, it seems better. So I guess the reason is the lack of feature matching in the background area of the outdoor scene, which leads to that part of the Gaussian not being well optimized.
Hello, Thanks for your nice work!
I want to know the hyper parameters settings that correspond to the results in your paper. I noticed that you set the iterations to 2000 and opacity reset interval to 200 in your released code. but I didn't get the same good results in T&T dataset as these you show in your paper, especially in outdoor scenes where the background is usually blurred without feature matching (Even in training views). I'm not sure if the opacity reset interval was set too small. Do you have some suggestions?
Here are some of my results, you can clearly find the render quality of background is bad ,which is different from the results in your papar.
In indoor scenes, it seems better. So I guess the reason is the lack of feature matching in the background area of the outdoor scene, which leads to that part of the Gaussian not being well optimized.
