OpenCV dnn cv.dnn.readNetFromONNX

[wllwel]

Hi！

Question

I trained the U2net model on my own custom dataset.
I converted ".pth" file to ONNX file. The ONNX model I created works fine.
I converted ONNX file to Quantize ONNX file. It cannot read the file "u2net_quant.onnx" that I created.
(I tried the official U2net model and the same problem occurred)

The error information:

[ERROR:0@3.332] global D:\a\opencv-python\opencv-python\opencv\modules\dnn\src\onnx\onnx_importer.cpp (909) cv::dnn::dnn4_v20211220::ONNXImporter::handleNode DNN/ONNX: ERROR during processing node with 1 inputs and 3 outputs: [DynamicQuantizeLinear]:(input_quantized) from domain='ai.onnx'
Traceback (most recent call last):
File "F:/mallPycharmProjects/projects/test/test.py", line 90, in
net = cv.dnn.readNetFromONNX('onnx/official/u2net_official_conF.quantU.onnx')
cv2.error: OpenCV(4.5.5) D:\a\opencv-python\opencv-python\opencv\modules\dnn\src\onnx\onnx_importer.cpp:928: error: (-2:Unspecified error) in function 'cv::dnn::dnn4_v20211220::ONNXImporter::handleNode'

Node [DynamicQuantizeLinear@ai.onnx]:(input_quantized) parse error: OpenCV(4.5.5) D:\a\opencv-python\opencv-python\opencv\modules\dnn\src\onnx\onnx_importer.cpp:508: error: (-5:Bad argument) Blob inputscale not found in const blobs in function 'cv::dnn::dnn4_v20211220::ONNXImporter::getBlob'

System Information

Python 3.8.8
OpenCV version: '4.5.5'
windows 7

[justinchuby]

This would be a good question for opencv. Thanks!