The effect of the garden-dataset is bad

[arvrschool]

the result after run_inference_inpainting is not good, the DSC07956_mask.ply is like this below:

and the result after finetune is also bad:

I think this problem may be due to the depth completion errors, can you give me some suggestions?
Thx!

[Hendery77]

Same question, have you solve it?

[Johanan528]

Sry, maybe you can further describe your problem? There are two possible reasons: 1. Our model uses disparity as input, which may have inf value. 2. Our model currently performs poorly with larger masks. For these two problems, we have trained a better and larger model. Using depth as input, it will be released in about a month.