search

chensjtu / GaussianObject

GaussianObject: High-Quality 3D Object Reconstruction from Four Views with Gaussian Splatting (SIGGRAPH Asia 2024, TOG)
887 stars 54 forks source link

Generating Masks from SegAnyGAussians #16

Open jjangsangy opened 7 months ago

jjangsangy commented 7 months ago

It seems in order to prepare your own dataset requires generating masks from SegAnyGAussians. However, looking through the repository and running their code, I was not able to find any way to generate the masks directory using their codebase.

Is there something I'm missing in the documentation, or do you guys have special code on top to generate the masks directory?

qcy27 commented 7 months ago

Same problem. Did you find a solution?

Thank you

GaussianObject commented 7 months ago

Thank you for your feedback regarding the use of SegAnyGAussians. It's important to note that SegAnyGAussians is primarily utilized for segmenting test views during our evaluation phase rather than being a requirement for practical application. For real-world sparse-input usage, we highly recommend adopting SegmentAnything.

GaussianObject commented 5 months ago

Hey there! We're developing a revolutionary pipeline that transforms 4 images into 3DGS with finedetailed and real-time rendering. All you need to do is click on an image (for SAM segmentation) and let GaussianObject do the rest! Stay tuned for more!