How to use MattingRefine?

[Willy-NotSpe]

I got the following error message after input the src and bgr imge.
Please help me, thanks.

Traceback (most recent call last):
File "./Matting/inference_images.py", line 131, in
pha, fgr, _, _, err, ref = model(src, bgr)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/app/Matting/model/model.py", line 189, in forward
pha, fgr, ref_sm = self.refiner(src, bgr, pha_sm, fgr_sm, err_sm, hid_sm)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/app/Matting/model/refiner.py", line 95, in forward
ref = self.select_refinement_regions(err)
File "/app/Matting/model/refiner.py", line 175, in select_refinement_regions
idx = err.topk(self.sample_pixels // 16, dim=1, sorted=False).indices
RuntimeError: invalid argument 5: k not in range for dimension at /pytorch/aten/src/THC/generic/THCTensorTopK.cu:26

[Willy-NotSpe]

Hi Rojin, I don't setup the data_path.py. Because I don't train the model for myself data, I just use the pre-train model Did that cause this problem? Thanks for your reply. best regards Rojin ***@***.***> 於 2022年3月25日 週五 下午5:56寫道：

…

Hi, did you setup data_path.py? best — Reply to this email directly, view it on GitHub <#172 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APLMKXTSZVECJIKZPHE2BADVBWETPANCNFSM5OGEGOGQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[grewanhassan]

No no, you are right, you don't need to setup data_path.py.
Can You upload your input image and the related background?

[Willy-NotSpe]

Sorry, I can't provide the data generated by the video.
Because the video was obtained from a private organization.

[grewanhassan]

Do You use inference_images?

[Willy-NotSpe]

Yes. I don't know why two different model types have the same steps in the first place.
"mattingbase" can be done, but another cannot.

[OoKyleoO]

HI I've got the same problem here. Any solutions yet?