Closed wing158 closed 4 months ago
Hi! :)
Our method and code are not limited to any resolution and number of views, but there will be differences in performance and inference speed for different settings of datasets.
Higher-resolution training images will result in clearer reconstruction results. A smaller number of training views can easily lead to overfitting of specific sparse views, which is an inherent problem of Gaussian Splatting.
请问demo里一般是800*800,200张图左右, 2K或4K的是否可以,points数里是否有限制?