Segmentation fault for cv2.dnn.readNetFromDarknet.forward() on macOS with 4.7.0.68 #23072
Description
System Information
OpenCV python version: 4.7.0.68
Operating System / Platform: macos-latest and macos-11
Python version: 3.11.1
It passes with 4.6.0.66 on all plattforms including macos.
It passes with 4.7.0.68 on linux and windows
Detailed description
Fatal Python error: Segmentation fault occurs for cv2.dnn.readNetFromDarknet.forward()
Steps to reproduce
print('cv2.imread')
img = cv2.imread('tests/data/test_googleauth_export.png')
yolo_v3_QR_detector = cv2.dnn.readNetFromDarknet(cfgFile='tests/data/qrcode-yolov3-tiny.cfg', darknetModel='tests/data/qrcode-yolov3-tiny_last.weights')
yolo_v3_QR_detector.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
output_layer_names = yolo_v3_QR_detector.getLayerNames()
output_layer_name = output_layer_names[yolo_v3_QR_detector.getUnconnectedOutLayers()[0] - 1]
print('cv2.dnn.blobFromImage')
_INPUT_SIZE = (416, 416)
_CONF_THRESHOLD = 0.5
blob = cv2.dnn.blobFromImage(img, 1 / 255, _INPUT_SIZE, swapRB=False, crop=False)
print('yolo.yolo_v3_QR_detector.setInput')
yolo_v3_QR_detector.setInput(blob=blob)
# Transform the image to blob and predict
print('yolo_v3_QR_detector.forward')
output = yolo_v3_QR_detector.forward(output_layer_name) # <========= Segmentation fault
print('Done')
Full code to reproduce including data can be found in the following branch: https://github.com/scito/extract_otp_secret_keys/tree/cv2_segfault_repro_1
Note: GitHub actions CI is enabled on this branch.
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
- I updated to the latest OpenCV version and the issue is still there
- There is reproducer code and related data files (videos, images, onnx, etc)