CHENG-DE FAN
CHENG-DE FAN
您好,請問您會提供render.py跟metrix.py讓我們eval最終結果嗎?
Thank you very much for your work. Could you please explain where 3DGS is used in your code? It looks different from typical 3DGS code.
Dear authors: Thanks for your great work! I hope to receive your guidance. Without using “SIBR Viewer”,how do I need to set so that get 3DGS ellipsoid.
您好,非常感謝您release code,想請問一下我在gaussian_model.py看到你在Densification的時候只考慮了canonical gaussian 的Densification,並沒有考慮deform後的gaussian的Densification,例如下面的連結的code只考慮了canonical gaussian 的scaling,這樣在Densification時是否不太合理呢? https://github.com/hustvl/4DGaussians/blob/master/scene/gaussian_model.py#L448
非常感謝您釋出您的code,想請問一下你們當初不對除了position以外的變數做learning rate schedule是有什麼原因嗎? 如下面這個函式只對gaussian的position 做update learning rate,沒有對其他變數例如scale、rotation做update learning rate。 https://github.com/hbb1/2d-gaussian-splatting/blob/main/scene/gaussian_model.py#L168
Hi 非常感謝你釋出的code,可以請問一下如何才能render出像你的paper圖片一樣的disk嗎
Can the surf_depth produced by render be used for depth supervision with the depth map generated by other pre-train models?
您好非常感謝您open source您的code,這是一篇很傑出的work。 想再這邊請問一下您知道如何不用SIBR viewer也可以visualize 3D Gaussian 橢球的方法嗎? 或者更改cuda code的哪個部分才可以做到呢?
I'd like to ask if, after saving the environment map with the code below and then loading it back, the environment map obtained is different from the original one.
Can I use a custom dataset, such as the BS-ERGB dataset, for inference?