WARNING: Could not fuse any points.

[ThePrecoder]

When I run on my own data,it occured that error.However,it was alright on your datasets.


I trained it use four gpus,so I guessed that is not the problem.and I altered the depth_H and W to my picture format.
By the way.Before the fusion step,It occured configuration problem and PatchMatch::Problem.I just dont know what happened.

[ThePrecoder]

and I found that when I used your provided dataset organized like the "own data",the fusion step is okay.Does it have some constraints on the data?

[XYZ-qiyh]

CUDA: out of memory.
please check your GPU memory consumption.

[weiyithu]

I also think it may be caused by GPU memory overflow. The resolution of your images are ?

[ThePrecoder]

3648 * 4864,93 pictures.run on four 2080Ti.

[ThePrecoder]

I also think it may be caused by .GPU memory overflow. The resolution of your images are ?

I thought the four gpus is enough to process this.The fact seems not,thanks a lot.But after fied this issue ,it's still went wrong(without CUDA error).

I changed the resolution of image to (912, 1216, 3),when it ran,the fusion step seems okay,.However,it still failed to run after codes.

[weiyithu]

How many picture names in train.txt ? The error comes from https://github.com/weiyithu/NerfingMVS/blob/b7f6590be69611c814822dc32e2220857c4d1dbf/utils/pose_utils.py#L48. It is caused by the misalignment between image numbers in train.txt and images.bin.

[ThePrecoder]

THANKS A LOT!I check the train.txt,it was my mistake.I did not carefully examine the contents of the document.It's okay now.