search

hbb1 / 2d-gaussian-splatting

[SIGGRAPH'24] 2D Gaussian Splatting for Geometrically Accurate Radiance Fields
https://surfsplatting.github.io
Other
1.94k stars 122 forks source link

Add Mask #52

Open li1u opened 4 months ago

li1u commented 4 months ago

Hello, I'm sorry to bother you. How do you remove the background from the image? I inputted the corresponding mask image, but the result he rendered still includes the background

li1u commented 4 months ago

To be precise, I want to remove the background information from the reconstructed 3D model

shenggedeqiang commented 3 months ago

same question

li1u commented 3 months ago

其实关键在于作者裁剪模型的时候不太对，你可以看一下gaussian_surfels的代码，把他的裁剪模型的部分弄到这个里面，我是成功了，效果挺好的

li1u commented 3 months ago

然后读取掩模数据的时候也需要注意下

hbb1 commented 3 months ago

I pack rgb with mask to be png images. The program will then treat the 4th channel as mask used for bounded TSDF fusion.

ch1998 commented 3 months ago

I pack rgb with mask to be png images. The program will then treat the 4th channel as mask used for bounded TSDF fusion.

In this way, can I directly use the RBGA image with the background removed to train?

AndreaCastiella commented 2 months ago

I pack rgb with mask to be png images. The program will then treat the 4th channel as mask used for bounded TSDF fusion.

In this way, can I directly use the RBGA image with the background removed to train?

Hello! Did you manage to train without background using rgba images?

hbb1 commented 2 months ago

I think it is trival to do so. Simply packing the rendered_rgb and render_alpha we get an RGBA image. Next, supervise it with the GT RGBA image.