MVS refine instruction

[Hsuan1079]

Dear AkiraHero,

I'm trying to implement the stage3 refinement part. I didn't see any code or ckpt that link to what you mentioned in the paper "MVS refinement". Cloud you please release the code or give me some advice how to reproduce the result on the paper?

The follow is my result and implement step. I try to use stage 3 only to refine the original scene. So first I encode the original tsdf to code1 and 2 to get the occupacy, and also use PatchVQGAN to get the original scene latent. Then only inference the stage3 by using the code1&2 with the computing mask. But I still get a bad result on the scannet scene. Cloud you please give me some advice or hint that how you did?