Inconsistent shape for ConcatLayer in function 'getMemoryShapes' while reading onnx model about opencv HOT 5 CLOSED

No error : you must simplify model :

onnxsim model(1).onnx a.sim.onnx

opencv onnx importer does not process dynamic shape. You have to substitue w and h

python -m onnxruntime.tools.make_dynamic_shape_fixed --dim_param w --dim_value 640 a.sim.onnx a1.onnx

python -m onnxruntime.tools.make_dynamic_shape_fixed --dim_param h --dim_value 480 a1.onnx a2.onnx

then
python

Python 3.10.10 (tags/v3.10.10:aad5f6a, Feb  7 2023, 17:20:36) [MSC v.1929 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2 as cv
>>> x = cv.dnn.readNet("a2.onnx")
>>> import numpy as np
>>> img =  np.zeros((1, 3, 480,640))
>>> x.setInput(img)
>>> w = x.forward()
>>>

from opencv.

This bug should be fixed if possible.

bug fixed

from opencv.

This bug should be fixed if possible. dexiened is used to replace hed_pretrained_bsds model for edge detection. Do we have other public onnx models without manual modification for edge detection? Use public model link will save our traffic.

from opencv.

This bug should be fixed if possible.

#19347 (comment)

from opencv.

This bug should be fixed if possible.

#19347 (comment)

OK, Thanks! Replaced model PR is crested #25515

from opencv.