Error with windows fast installation

[yohlimem]

Traceback (most recent call last):
File "D:\FasterLivePortrait-windows-20240811\run.py", line 86, in
dri_crop = cv2.resize(dri_crop, (512, 512))
cv2.error: OpenCV(4.10.0) D:\a\opencv-python\opencv-python\opencv\modules\imgproc\src\resize.cpp:4152: error: (-215:Assertion failed) !ssize.empty() in function 'cv::resize'

nvcc --version:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Jun_13_19:42:34_Pacific_Daylight_Time_2023
Cuda compilation tools, release 12.2, V12.2.91
Build cuda_12.2.r12.2/compiler.32965470_0

I think the model returns None but I'm not 100% sure. Any help would be great.

[mmattyg]

I'm getting a similiar problem. It looks like the onnx models are returning empty (None) results.

Any good way to debug this?

[yohlimem]

It looks like if I set the cfg file to
configs/onnx_mp_infer.yaml
The media pip config is not working, and after that, NOTHING is working anymore (even if going back to regular onnx_infer.yaml)
It just breaks everything somehow.

Anyone has any idea what could be going wrong? Only workaround I found is reinstalling everything..

[yohlimem]

update: it's still working with a driving video, just not with the camera as a driving source
still weird..

[leetraman822]

👋Hello, the cv:resize error could be due to your camera not working, being occupied, or not detecting a face within it. Before running run.py, ensure there is a clear, visible face in the camera's view. This significantly reduces the occurrence of cv:resize errors. In rare cases, despite meeting all previous conditions, the cv:resize error might still occasionally appear. This could be because, during face detection, the face was not successfully recognized in that particular frame due to lighting or angle issues. Nonetheless, this approach is generally effective.🙂

[warmshao]

👋Hello, the cv:resize error could be due to your camera not working, being occupied, or not detecting a face within it. Before running run.py, ensure there is a clear, visible face in the camera's view. This significantly reduces the occurrence of cv:resize errors. In rare cases, despite meeting all previous conditions, the cv:resize error might still occasionally appear. This could be because, during face detection, the face was not successfully recognized in that particular frame due to lighting or angle issues. Nonetheless, this approach is generally effective.🙂

Thank you, I will handle this kind of error in the code update later.